Académique Documents
Professionnel Documents
Culture Documents
MECHANICS
ELSEVIER
PHYSICS
REPORTS
Contents:
1. Introduction
2. Hamiltonian formulation of supersymmetric
quantum mechanics
2.1. State space structure of the SUSY harmonic
oscillator
2.2. Broken supersymmetry
3. Factorization and the hierarchy of Hamiltonians
4. Shape invariance and solvable potentials
4.1. General formulas for bound state spectrum,
wave functions and S-matrix
4.2. Shape invariance in more than one step
4.3. Strategies for categorizing shape invariant
potentials
4.4. Shape invariance and noncentral solvable
potentials
4.5. Shape invariance and 3-body solvable
potentials
5. Operator transforms - new solvable potentials
from old
5.1. Natanzon potentials
5.2. Generalizations of Ginocchio and Natanzon
potentials
6. Supersymmetric WKB approximation
271
276
283
285
288
290
291
292
294
307
308
311
314
316
319
0370.1573/95/!$29.00
@ 1995 Elsevier Science B.V. All rights reserved
.SSDIO370-1573(94)00080-M
319
321
322
323
324
325
327
329
330
336
338
338
341
342
346
348
350
351
352
354
355
356
12.Singular superpotentials
12.i. General formalism, negative energy states
and breakdown of the degeneracy theorem
12.2. Bound states in the continuum
13.Parasupersymmetric quantum mechanics and
beyond
13.1. Parasupersymmetric quantum mechanics
13.2. Orthosupersymmetric quantum mechanics
14.0mitted topics
References
269
359
360
365
371
372
376
378
380
Abstract
In the past ten years, the ideas of supersymmetry have been profitably applied to many nonrelativistic quantum
mechanical problems. In particular, there is now a much deeper understanding
of why certain potentials are
analytically
solvable and an array of powerful new approximation
methods for handling potentials which
are not exactly solvable. In this report, we review the theoretical formulation of supersymmetric
quantum
mechanics and discuss many applications. Exactly solvable potentials can be understood in terms of a few
basic ideas which include supersymmetric
partner potentials, shape invariance and operator transformations.
Familiar solvable potentials all have the property of shape invariance. We describe new exactly solvable shape
invariant potentials which include the recently discovered self-similar potentials as a special case. The connection
between inverse scattering, isospectral potentials and supersymmetric quantum mechanics is discussed and multisoliton solutions of the KdV equation are constructed. Approximation
methods are also discussed within the
framework of supersymmetric
quantum mechanics and in particular it is shown that a supersymmetry
inspired
WKB approximation
is exact for a class of shape invariant potentials. Supersymmetry
ideas give particularly
nice results for the tunneling rate in a double well potential and for improving large N expansions. We also
discuss the problem of a charged Dirac particle in an external magnetic field and other potentials in terms of
supersymmetric
quantum mechanics, Finally, we discuss structures more general than supersymmetric quantum
mechanics such as parasupersymmetric
quantum mechanics in which there is a symmetry between a boson and
a para-fermion of order p.
270
1. Introduction
Physicists have long strived to obtain a unified description of all basic interactions of nature, i.e.
strong, electroweak, and gravitational interactions. Several ambitious attempts have been made in the
last two decades, and it is now widely felt that supersymmetry (SUSY) is a necessary ingredient in
any unifying approach. SUSY relates bosonic and fermionic degrees of freedom and has the virtue
of taming ultraviolet divergences. It was discovered in 1971 by Gelfand and Likhtman [ 11, Ramond
[ 21 and Neveu and Schwartz [ 31 and later was rediscovered by several groups [ 4-61. The algebra
involved in SUSY is a graded Lie algebra which closes under a combination of commutation and
anti-commutation
relations. It was first introduced in the context of the string models to unify the
bosonic and the fermionic sectors. It was later shown by Wess and Zumino [5] how to construct
a 3 + 1 dimensional field theory which was invariant under this symmetry and which had very
interesting properties such as a softening of ultraviolet divergences as well as having paired fermionic
and bosonic degrees of freedom. For particle theorists, SUSY offered a possible way of unifying
space-time and internal symmetries of the S-matrix which avoided the no-go theorem of Coleman
and Mandula [7] which was based on the assumption of a Lie algebraic realization of symmetries
(graded Lie algebras being unfamiliar to particle theorists at the time of the proof of the no-go
theorem). Gravity was generalized by incorporating SUSY to a theory called supergravity [ 8,9]. In
such theories, Einsteins general theory of relativity turns out to be a necessary consequence of a
local gauged SUSY. Thus, local SUSY theories provide a natural framework for the unification of
gravity with the other fundamental interactions of nature.
Despite the beauty of all these unified theories, there has so far been no experimental evidence
of SUSY being realized in nature. One of the important predictions of unbroken SUSY theories is
the existence of SUSY partners of quarks, leptons and gauge bosons which have the same masses as
their SUSY counterparts. The fact that no such particles have been seen implies that SUSY must be
spontaneously broken. One hopes that the scale of this breaking is of the order of the electroweak
scale of 100 GeV in order that it can explain the hierarchy problem of mass differences. This leads
to a conceptual problem since the natural scale of symmetry breaking is the gravitational or Planck
scale which is of the order of 1019 GeV. Various schemes have been invented to try to resolve the
hierarchy problem, including the idea of non-perturbative breaking of SUSY. It was in the context of
this question that SUSY was first studied in the simplest case of SUSY quantum mechanics (SUSY
QM) by Witten [ lo] and Cooper and Freedman [ 111. In a subsequent paper, a topological index
was introduced (the Witten index) by Witten [ 121 for studying SUSY breaking and several people
studied the possibility that instantons provide the non-perturbative mechanism for SUSY breaking.
In the work of Bender et al. [ 131 a new critical index was introduced to study non-perturbatively
the breakdown of SUSY in a lattice regulated theory. Thus, in the early days, SUSY was studied in
quantum mechanics as a testing ground for the non-perturbative methods of seeing SUSY breaking
in field theory.
Once people started studying various aspects of SUSY QM, it was soon clear that this field was
interesting in its own right, not just as a model for testing field theory methods. It was realized that
SUSY gives insight into the factorization method of Infeld and Hull [ 141 which was the first method
to categorize the analytically solvable potential problems. Gradually a whole technology was evolved
based on SUSY to understand the solvable potential problems. One purpose of this article is to review
some of the major developments in this area.
271
212
a wide class of solvable potentials by considering six different forms of factorization. It turns out
that the factorization method as well as the methods of SUSY QM including the concept of shape
invariance (with translation of parameters), are both reformulations [ 501 of Riccatis idea of using
the equivalence between the solutions of the Riccati equation and a related second order linear
differential equation. This method was supposedly used for the first time by Bernoulli and the history
is discussed in detail by Stahlhofen [ 5 11.
The general problem of the classification of SIPS has not yet been solved. A partial classification of
the SIPS involving a translation of parameters was done by Cooper et al. [ 52,531. It turns out that in
this case one only gets all the standard analytically solvable potentials contained in the list given by
Dutt et al. [ 541 except for one which was later pointed out by Levai [ 551, The connection between
SUSY, shape invariance and solvable potentials [56,57] is also discussed in the paper of Cooper
et al. [52] where these authors show that shape invariance even though sufficient, is not necessary
for exact solvability. Recently, a new class of SIPS has been discovered which involves a scaling of
parameters [ 581. These new potentials as well as multi-step SIPS [ 591 have been studied, and their
connection with self-similar potentials as well as with q-deformations has been explored [ 60-631.
In yet another development, several people showed that SUSY QM offers a simple way of obtaining
isospectral potentials by using either the Darboux [64] or Abraham-Moses [65] or Pursey [66]
techniques, thereby offering glimpses of the deep connection between the methods of the inverse
quantum scattering [67], and SUSY QM [ 68-721. The intimate connection between the soliton
solutions of the KdV hierarchy and SUSY QM was also brought out at this time [ 73-761,
Approximate methods based on SUSY QM have also been developed. Three of the notable ones
are the 1/N expansion within SUSY QM [ 771, 6 expansion for the superpotential [ 781 and a SUSY
inspired WKB approximation (SWKB) in quantum mechanics for the case of unbroken SUSY [ 7981 1. It turns out that the SWKB approximation preserves the exact SUSY relations between the
energy eigenvalues as well as the scattering amplitudes of the partner potentials [ 821. Further, it is
not only exact for large n (as any WKB approximation is) but by construction it is also exact for the
ground state of VI (x) . Besides it has been proved [45] that the lowest order SWKB approximation
necessarily gives the exact spectra for all SIP (with translation). Subsequently a systematic higher
order SWKB expansion has been developed and it has been explicitly shown that to 0(ti6) all the
higher order corrections are zero for these SIP [ 831. This has subsequently been generalised to all
orders in ti [ 84,851. Energy eigenvalue spectrum has also been obtained for several non-SIP [ 86-891
and it turns out that in many of the cases the SWKB does better than the usual WKB approximation.
Based on a study of these and other examples, it has been suggested that shape invariance is not only
sufficient but perhaps necessary for the lowest order SWKB to give the exact bound state spectra
[ 901. Some attempts have also been made to obtain the bound state eigenfunctions within the SWKB
formalism [ 91-931.
Recently, Inomata and Junker [94] have derived the lowest order SWKB quantization condition
(BSWKB) in case SUSY is broken. It has recently been shown that for the cases of shape invariant
three dimensional oscillator as well as for Pbschl-Teller I and II potentials with broken SUSY, this
lowest order BSWKB calculation gives the exact spectrum [ 95,971. Recently, Dutt et al. [ 961 have
also developed a systematic higher order BSWKB expansion and using it have shown that in all the
three (shape invariant) cases, the higher order corrections to O(ti6) are zero. Further, the energy
eigenvalue spectrum has also been obtained in the case of several non-SIP and it turns out that in
many cases BSWKB does as well as (if not better than) the usual WKB approximation [ 961.
273
Recursion relations between the propagators pertaining to the SUSY partner potentials have been
obtained and explicit expressions for propagators of several SIP have been obtained [ 98,991.
Several aspects of the Dirac equation have also been studied within SUSY QM formalism [ IOO1021. In particular, it has been shown using the results of SUSY QM and shape invariance that
whenever there is an analytically solvable Schrijdinger problem in l-dimensional
QM then there
always exists a corresponding Dirac problem with scalar interaction in 1+1 dimensions which is
also analytically solvable. Further, it has been shown that there is always SUSY for massless Dirac
equation in two as well as in four Euclidean dimensions. The celebrated problem of the Dirac particle
in a Coulomb field has also been solved algebraically by using the concepts of SUSY and shape
invariance [ 1031. The SUSY of the Dirac electron in the field of a magnetic monopole has also been
studied [ 104,105]. Also, the classic calculation of Schwinger on pair production from strong fields
can be dramatically simplified by exploiting SUSY.
The formalism of SUSY QM has also been recently extended and models for parasupersymmetric
QM [ 106-1081 as well as orthosupersymmetric
QM [ 1091 of arbitrary order have been written down.
The question of singular superpotentials has also been discussed in some detail within SUSY QM
formalism [ 110-l 141. Very recently it has been shown that SUSY QM offers a systematic method
[ 1151 for constructing bound states in the continuum [ 116-1181.
As is clear from this (subjective) review of the field, several aspects of SUSY QM have been
explored in great detail in the last ten years and it is almost impossible to cover all these topics and
do proper justice to them. We have therefore, decided not to pretend to be objective but cover only
those topics which we believe to be important and which we believe have not so far been discussed in
great detail in other review articles. We have, however, included in Section 14 a list of the important
topics missed in this review and given some references so that the intersested reader can trace back
and study these topics further. We have been fortunate in the sense that review articles already exist
in this field where several of these missing topics have been discussed [ 119-122,54,123-125,100].
We must also apologize to several authors whose work may not have been adequately quoted in this
review article in spite of our best attempts.
The plan of the article is the following: In Section 2, we discuss the Hamiltonian formalism of
SUSY QM. We have deliberately kept this section at a pedagogical level so that a graduate student
should be able to understand and work out all the essential details. The SUSY algebra is given and the
connection between the energy eigenvalues, the eigenfunctions and the S-matrics of the two SUSY
partner Hamiltonians are derived. The question of unbroken vs. broken SUSY is also introduced at
a pedagogical level using polynomial potentials of different parity and the essential ideas of partner
potentials are illustrated using the example of a one dimensional infinite square well. The ideas of
SUSY are made more explicit through the example of one dimensional SUSY harmonic oscillator.
In Section 3, we discuss the connection between SUSY QM and factorization and show how one can
always construct a hierarchy of p > 1 Hamiltonians with known energy eigenvalues, eigenfunctions
and S-matrices by starting from any given Hamiltonian with p bound states whose energy eigenvalues,
eigenfunctions and S-matrices (reflection coefficient R and transmission coefficient T) are known.
Section 4 is in a sense the heart of the article. We first show that if the SUSY partner potentials
satisfy an integrability condition called shape invariance then the energy eigenvalues, the eigenfunctions and the S-matrices for these potentials can be obtained algebraically. We then discuss satisfying
the condition of shape invariance with translations and show that in this case the classification of
SIP can be done and the resultant list of solvable potentials include essentially all the popular ones
214
that are included in the standard QM textbooks. Further, we discuss the newly discovered one and
multi-step shape invariant potentials when the partner potentials are related by a change of parameter
of a scaling rather than translation type. It turns out that in most of these cases the resultant potentials are reflectionless and contain an infinte number of bound states. Explicit expressions for the
energy eigenvalues, the eigenfunctions and the transmission coefficients are obtained in various cases.
It is further shown that the recently discovered self-similar potentials which also statisfy q-SUSY,
constitute a special case of the SIP. Finally, we show that a wide class of noncentral but separable
potential problems are also algebraically solvable by using the results obtained for the SIP [ 1311.
As a by product, exact solutions of a number of three body problems in one dimension are obtained
analytically [ 1321.
In Section 5, we discuss the solvable but non-shape invariant Natanzon and Ginocchio potentials
and show that using the ideas of SUSY QM, shape invariance and operator transformations, their
spectrum can be obtained algebraically. We also show that the Natanzon potentials are not the most
general solvable potentials in nonrelativistic QM.
Section 6 is devoted to a discussion of the SUSY inspired WKB approximation
(SWKB) in
quantum mechanics both when SUSY is unbroken and when it is spontaneously broken. In the
unbroken case, we first develop a systematic higher order ti -expansion for the energy eigenvalues
and then show that for the SIP with translation, the lowest order term in the &expansion gives the
exact bound state spectrum. We also show here that even for many of the non-SIP, the SWKB does
as well as if not better than the WKB approximation. We then discuss the broken SUSY case and in
that case too we develop a systematic &expansion (BSWKB) for the energy eigenvalues. We show
that even in the broken case the lowest order BSWKB gives the exact bound state spectrum for the
SIP with translation.
Section 7 contains a description of how SUSY QM can be used to construct multiparameter families
of isospectral and strictly isospectral potentials. As an illustration we give plots of the one continuous
parameter family of isospectral potentials corresponding to the one-dimensional harmonic oscillator.
From here we are immediately able to construct the two as well as multisoliton solutions of the KdV
equation.
In Section 8, we discuss more formal aspects of SUSY QM. In particular, we discuss the path
integral formulation of SUSY QM as well as various subtleties associated with the Witten index [ 121
d = (- 1) F. We also discuss in some detail the connection of SUSY QM with classical stochastic
processes and discuss how one can develop a systematic strong coupling and 8 -expansion for the
Langevin equation.
Section 9 contains a description of several approximation schemes like the variational method,
the S-expansion, large-h expansion, energy splitting in double well potentials within the SUSY QM
framework.
In Section 10, we discuss the question of SUSY QM in higher dimensions. In particular, we discuss
the important problem of a charged particle in a magnetic field (Pauli equation) in two dimensions
and show that there is always a SUSY in the problem so long as the gyromagnetic ratio is 2.
In Section 11, we show that there is always a SUSY in the case of massless Dirac equation in
two or four Euclidean dimensions in the background of external electromagnetic
fields. Using the
results of SUSY QM we then list a number of problems with nonuniform magnetic field which can
be solved analytically. We also show here that whenever a Schrodinger problem in l-dimensional
QM is analytically solvable, then one can always obtain an exact solution of a corresponding Dirac
275
problem with the scalar coupling. We also show how the calculation of the fermion propagator in an
external field can be simplified by exploiting SUSY.
Section 12 contains a comprehensive discussion of the general problem of singular superpotentials,
explicit breaking of SUSY, negative energy states and unpaired positive energy eigenstates. We also
show here how to construct bound states in the continuum within the formalism of SUSY QM.
Quantum mechanical models relating bosons and parafermions of order p are described in Section
13. It is shown that such models encompass p supersymmetries. Various consequences of such models
are discussed including the connection with the hierarchy of Hamiltonians as well as with strictly
isospectral potentials. We also discuss a quantum mechanical model where instead there is a symmetry
between a boson and an orthofermion of order p.
Finally, in Section 14, we give a list of topics related to SUSY QM which we have not discussed
and provide some references for each of these topics.
h* d2i,bo
so that
h2 @l(x)
v,(x) = -~
2m ccl0W
(2)
.
This allows a global reconstruction of the potential V, (x) from a knowledge of its ground state wave
function which has no nodes (we will discuss the case of using the excited wave functions later
in Section 12). Once we realize this, it is now very simple to factorize the Hamiltonian using the
following ansatz:
H1 = A+A
(3)
where
A=-&+W(n),
A=~$+W(x).
(4)
v&i
w*(x) - -W(x).
(3
276
This equation is the well-known Riccati equation. The quantity W(x) is generally referred to as the
superpotential in SUSY QM literature. The solution for W(x) in terms of the ground state wave
function is
(6)
This solution is obtained by recognizing that once we satisfy A& = 0, we automatically have a
solution to Hl@a = A+A& = 0.
The next step in constructing the SUSY theory related to the original Hamiltonian HI is to define
the operator H2 = AA+ obtained by reversing the order of A and A+. A little simplification shows that
the operator Hz is in fact a Hamiltonian corresponding to a new potential V2(x) .
v,(x) = W2(x)+
The potentials V,(x) and V,(x) are known as supersymmetric partner potentials.
As we shall see, the energy eigenvalues, the wave functions and the S-matrices of HI and H2 are
related. To that end notice that the energy eigenvalues of both HI and H2 are positive semi-definite
(E(1*2)
>
n
- 0) . For n > 0, the Schrtidinger equation for HI
H&A) = AtA@;) = E,I)$;)
(8)
implies
H2(At,b(I))
= AA+A#*
n = E;(A#).
n
(9)
= AA+&2
= E;2)$;2)
(10)
implies
HI ( A+$c2)
n = Ei2( A++c2)
n = A+AA+@2
-
(111
From Eqs. (8)-( 11) and the fact that E,,(I) - 0 , it is clear that the eigenvalues and eigenfunctions of
the two Hamiltonians HI and HZ are related by (n = 0, 1,2, . . .)
E(2) = E;?,,
n
t,b; = [ E$]
E; = 0,
-At,b$,
(12)
(13)
(14)
Notice that if @,$\ ( @A2))of HI ( H2) is normalized then the wave function $A) (@,$\ > in Eqs. ( 13)
and ( 14) is also normalized. Further, the operator A (At) not only converts an eigenfunction of HI
( H2) into an eigenfunction of H2( HI) with the same energy, but it also destroys (creates) an extra
node in the eigenfunction. Since the ground state wave function of HI is annihilated by the operator
A, this state has no SUSY partner. Thus the picture we get is that knowing all the eigenfunctions
of HI we can determine the eigenftmctions of HZ using the operator A, and vice versa using At we
F: Cooper et al/Physics
11)
277
(2)
El
EO
(1 1
EO
V,(x)
V,(x)
Fig. 2.1. The energy levels of two supersymmetric partner potentials VI and b. The figure corresponds to unbroken SUSY.
The energy levels are degenerate except that 6 has an extra state at zero energy E,,() = 0 . The action of the operators A
and At in connecting eigenfunctions is shown.
can reconstruct all the eigenfunctions of HI from those of Hz except for the ground state. This is
illustrated in Fig. 2.1.
The underlying reason for the degeneracy of the spectra of HI and H2 can be understood most
easily from the properties of the SUSY algebra. That is we can consider a matrix SUSY Hamiltonian
of the form
(15)
which contains both HI and Hz. This matrix Hamiltonian is part of a closed algebra which contains
both bosonic and fermionic operators with commutation and anti-commutation relations. We consider
the operators
(16)
(17)
in conjunction
with H. The following commutation and anticommutation relation s then describe the
closed superalgebra sZ( 1 / 1) :
(12, Q+) = fL
Cl@
The fact that the supercharges Q and Qt commute with H is responsible for the degeneracy. The
operators Q and Qt can be interpreted as operators which change bosonic degrees of freedom into
fermionic ones and vice versa. This will be elaborated further below using the example of the SUSY
harmonic oscillator.
Let us look at a well known potential, namely the infinite square well and determine its SUSY
partner potential. Consider a particle of mass m in an infinite square well potential of width L.
278
V(x)
=o,
O<X<L,
=oO,
--oo<x<o,
x > L.
(19)
= (2/L)/*sin(rx/L),
0 5 x _<L,
(20)
(21)
(ii(l)
n = (2/L)/*sin
O<x<L.
(22)
The superpotential for this problem is readily obtained using Eq. (6)
W(x) = --
cot@x/L)
&:
(23)
= gg
(24)
The wave functions for H2 are obtained by applying the operator A to the wave functions of HI. In
particular we find that
&* c( sin*(?rx/L),
$, c( sin(?rx/L) sin(29rx/L).
(25)
Thus we have shown using SUSY that two rather different potentials corresponding to HI and
H2 have exactly the same spectra except for the fact that H2 has one fewer bound state. In Fig.
2.2 we show the supersymmetric partner potentials V, and V2 and the first few eigenfunctions. For
convenience we have chosen L = T and ti = 2m = 1.
Supersymmetry also allows one to relate. the reflection and transmission coefficients in situations
where the two partner potentials have continuum spectra. In order for scattering to take place in both
of the partner potentials, it is necessary that the potentials V,,2 are finite as x --f -c0 or as x --f +oo
or both. Define:
W(x --+ 500) = w*.
(26)
Then
52
-+
w:
as
x+&c.
(27)
Let us consider an incident plane wave eikx of energy E coming from the direction x -+ --oo. As
a result of scattering from the potentials Vr,2(x) one would obtain transmitted waves 7i,2( k) 8 and
reflected waves RI,2( k) e- ikx. Thus we have
V,(x)
71
0
X
279
2 cosec2 x
7r
0
X
Fig. 2.2. The infinite square well potential V = 0 of width L = ?r and its supersymmetric partner potential 2cosecx in units
ti = 2m = 1. The ground state of the infinite square well has energy 1. Note the degenerate higher energy levels at energies
22,32,42,. . .
q!~(,~)(k, ,x --t -00)
$F
(k, x -+ +co)
-+ eik. + R1,2e-ikx,
+ 7i2eikfx.
(28)
SUSY connects continuum wave functions of Hi and H2 having the same energy analogously
happens in the discrete spectrum. Thus we have the relationships:
eikx
Rle-ikx
N[
(-ik
to what
where N is an overall
eliminating N, we find:
normalization
(29)
constant.
On equating
and
(30)
where k and k are given by
k = (E - Wf)2,
k = (E - W:)12.
(31)
280
E Cooper
et al./Physics
267-385
remarks that if one of the partner potentials is a constant potential (i.e. a free
partner will be of necessity reflectionless. In this way we can understand the
of the form V(X) = Asech*ax which play a critical role in understanding the
KdV hierarchy. Let us consider the superpotential
= A tanhax.
(32)
(33)
Then it is easily shown [ 1261 that the reduced radial wave function R satisfies the one-dimensional
Schrijdinger equation (0 < r < 00)
ii2 d'+(r)
--2m
dr*
+ IV(r)
Z(E+ l)@
2mr2
1$(r)
= E@(r)
cCl(r,I) -+
_(_l)le-ikr],
&[S(k)e.
(35)
(36)
where S1 is the scattering function for the Zth partial wave. i.e. S(k) = &I(~) and 6 is the phase shift.
For this case we find the relations:
(37)
Here W+ = W(r + co).
We thus have seen that when Hi contained a known ground state wave function then we could
factorize the Hamiltonian and find a SUSY partner Hamiltonian Hz, Now let us consider the converse
problem. Suppose we are given a superpotential W(x). In this case there are two possibilities. The
candidate ground state wave function is the ground state for Hi (or Hz) and can be obtained from:
At&)(x)
At+A2)(x)
=0
=0
&(x)
281
= Nexp
+ t,bi2(x) = Nexp
(38)
By convention, we shall always choose W in such a way that amongst Hi, H2 only H, (if at all)
will have a normalizable zero energy ground state eigenfunction. This is ensured by choosing W such
that W(x) is positive (negative) for large positive (negative) X. This defines HI to have fermion
number zero in our later formal treatment of SUSY.
If there are no normalizable solutions of this form, then HI does not have a zero eigenvalue and
SUSY is broken. Let us now be more precise. A symmetry of the Hamiltonian (or Lagrangian) can
be spontaneously broken if the lowest energy solution does not respect that symmetry, as for example
in a ferromagnet, where rotational invariance of the Hamiltonian is broken by the ground state. We
can define the ground state in our system by a two dimensional column vector:
(40)
Thus we have immediately from Eq. ( 18 ) that the ground state energy must be zero in this case.
For all the cases we discussed previously, the ground state energy was indeed zero and hence the
ground state wave function for the matrix Hamiltonian can be written:
(41)
where $$(x) is given by Eq. (38).
If we consider superpotentials of the form
W(X) =gP,
(42)
then for n odd and g positive one always has a normalizable ground state wave function. However for
the case IZeven and g arbitrary, then there is no normalizable ground state wave function. In general
when one has a superpotential W(x) so that neither Q nor Qt annihilates the ground state as given
by Eq. (39) then SUSY is broken and the potentials V, and b have degenerate positive ground state
energies. Stated another way, if the ground state energy of the matrix Hamiltonian is non zero then
SUSY is broken. For the case of broken SUSY the operators A and At no longer change the number
of nodes and there is a l-l pairing of all the eigenstates of HI and Hz. The precise relations that one
now obtains are:
E(2)
= E(l)
>
n=0,1,2,...
(43)
$:2) = &]-i&i),
(44)
(45)
= [E;2)]-1/2At@;2).
282
while the relationship between the scattering amplitudes is still given by Eqs. (30) or
breaking of SUSY can be described by a topological quantum number called the Witten
which we will discuss later. Let us however remember that in general if the sign of W(x)
as we approach infinity from the positive and the negative sides, then SUSY is unbroken,
the other cases it is always broken.
(37). The
index [ 121
is opposite
whereas in
+ tmo2q2.
(46)
coordinates
12
If=
Htiw,
q=
&
x,
P = (2mlio)f2p.
(47)
Then
H= (p2+
$x2),
[x,p]
=i.
Now introduce
a=(G+ip),
a+=(;--ip).
(49)
Then
[a,~+]
=1,
N = ata,
[N,a]
=-a,
H=N+&
[N,a+] =a+,
(50)
for solving the harmonic oscillator potential is to define the ground
(51)
which leads to a first order differential equation for the ground state wave function. The II particle
state (which is the nth excited wave function in the coordinate representation)
is then given by:
(52)
where we have used the subscript b to refer to the boson sector as distinct from the fermions we will
introduce below. For the case of the SUSY harmonic oscillator one can rewrite the operators Q ( Q +)
283
as a product of the bosonic operator a and the fermionic operator $. Namely we write Q = a$+ and
Q+ = a+$ where the matrix fermionic creation and annihilation operators are defined via:
I
=a-=
001
+=u+=
:,
(53)
@+
[1
(54)
Thus @ and $+ obey the usual algebra of the fermionic creation and annihilation
(fit&>
= 1,
{(Cr+?(Fl+}
= {@,fi} =O,
=(73
operators, namely,
(55)
relation:
(56)
(--g+;)+,,,.
H=QQ++Q+Q=
(57)
The effect of the last term is to remove the zero point energy.
The state vector can be thought of as a matrix in the Schrodinger picture or as the state jnb, nf) in
this Fock space picture. Since the fermionic creation and annihilation operators obey anti-commutation
relations hence the fermion number is either zero or one. As stated before, we will choose the ground
state of HI to have zero fermion number. Then we can introduce the fermion number operator
nF=-=1 -
CT3
l-
w4+1
2
(58)
= I& + 1, y),
fit%, q) = 1% nf - l),
et 1% q)
= (% nf + 1).
(59)
We now see that the operator Q+ = -ia@+ has the property of changing a boson into a fermion
without changing the energy of the state. This is the boson-fermion degeneracy characteristic of all
SUSY theories.
For the general case of SUSY QM, the operator a gets replaced by A in the definition of Q, Q+,
i.e. one writes Q = Afit and Q+ = A+t,b. The effect of Q and Q+ are now to relate the wave functions
of HI and Hz which have fermion number zero and one respectively but now there is no simple Fock
space description in the bosonic sector because the interactions are non-linear. Thus in the general
case, we can rewrite the SUSY Hamiltonian in the form
284
formulation
of SUSY QM in Section
H = {Q+, Q}.
Witten [ 121 proposed an index to determine whether SUSY is broken in supersymmetric
The index is defined by
n = Tr(-l)F,
field theories.
(61)
where the trace is over all the bound states and continuum states of the super-Hamiltonian. For SUSY
QM, the fermion number IZ~3 F is defined by i [ 1 - u3] and we can represent (-1) by the matrix
g3. If we write the eigenstates of H as the vector:
(62)
then the Z!Zcorresponds to the eigenvalues of ( - 1) F being f 1. For our conventions the eigenvalue + 1
corresponds to H, and the bosonic sector and the eigenvalue -1 corresponds to H2 and the fermionic
sector. Since the bound states of HI and H2 are paired, except for the case of unbroken SUSY where
there is an extra state in the bosonic sector with E = 0 we expect for the quantum mechanics situation
that n = 0 for broken SUSY and n = 1 for unbroken SUSY. In the general field theory case, Witten
gives arguments that in general the index measures N, (E = 0) - N_ (E = 0) which is the difference
n between the number of Bose states and Fermi states of zero energy. In field theories the Witten
index needs to be regulated to be well defined so that one considers instead
n(p)
= Tr(-l)Fe-PH,
(63)
(64)
(65)
As we said before, if this candidate ground state wave function does not fall off fast enough at
foe then Q does not annihilate the vacuum and SUSY is spontaneously broken. Let us show using
285
a trivial calculation that for two simple polynomial potentials the Witten index does indeed provide
the correct answer to the question of SUSY breaking. Let us consider
n(p)
Tra3
;;I
e-P[P2/2+w2/2-c+Sw(X)/2.
(66)
Expanding the term proportional to us in the exponent and taking the trace we obtain
A(P) = / [F]
e-prp2/2+WZ/2sinh(
/?W(x) /2),
(67)
We are interested in the regulated index as /? tends to 0, so that practically we need to evaluate
n(p)
= J [F]
f?-~p22+~221(pw(X)/2).
(68)
If we directly evaluate this integral for any potential of the form W(x) = gx2*+, which leads to a
normalizable ground state wave function, then all the integrals are gamma functions and we explictly
obtain A = 1. If instead W(x) = gx2 so that the candidate ground state wave function is not
normalizable then the integrand becomes an odd function of x and therefore vanishes. Thus we see
for these simple cases that the Witten index immediately coincides with the direct method available
in the quantum mechanics case.
Next let us discuss a favorite type of regularization scheme for field theory - namely the heat
kernel method. (Later we will discuss a path integral formulation for the regulated Witten index).
Following Akhoury and Comtet [ 161 one defines the heat kernels K* (x, y; ,B) which satisfy
-- -$ +
d
d/J
W2 7 W
K* = 0.
(69)
= /MK+(
X,-G/~) - K-(x,x;P)l,
or
A(p)
- p_(E)),
(71)
where p+ corresponds to the density of states. What Akhoury and Comtet were able to show, was
that in cases when W(x) went to different constants at plus and minus infinity, then the density of
states factors for the continuum did not cancel and that A(,@ could depend on p and be fractional
at p = 0. We refer the interested reader to the original paper for further details.
E Cooper et al/Physics
286
Another non-perturbative method for studying SUSY breaking in field theory is to explicitly break
SUSY by placing the theory on a lattice and either evaluating the path integral numerically or via some
lattice non weak- perturbative method such as the strong coupling (or high temperature) expansion.
This method was studied in detail in [ 13,211 and we will summarize the results here. The basic idea
is to introduce a new parameter, namely the lattice spacing a. This parameter explicitly breaks SUSY
so that the ground state energy of the system will no longer be zero, even in the unbroken case. One
hopes that as the lattice spacing is taken to zero then the ground state energy will go to zero as a
power of the lattice spacing if SUSY is unbroken, that is we expect
&)(a) = cay,
(72)
where y is a critical index which if greater than zero should be easy to measure in a Monte Carlo
calculation. Measuring the ground state energy at two different lattice spacings one studies:
y=ln---
&(a)
ln<
Eo(a) I
(73)
in the limit a and a -+ 0. The case of broken SUSY is more difficult because then we expect y = 0
which is a hard measurement to make numerically. In this latter case it is easier to directly measure
the ground state energy and show that it remains non-zero as one takes the lattice spacing to zero.
To see how this works in quantum mechanics one can do a lattice strong coupling expansion of the
Langevin equation which allows one to determine the ground state wave function of Ht as we shall
show later.
For the superpotential W(X) = gx3, we expect to find a positive critical index since here the
candidate ground state wave function is proportional to e-gx4/4 and is normalizable. The ground state
expectation values of X for the Hamiltonian Hr can be determined by first solving the Langevin
equation
$ +W(x)
= r](t)
x(7(t)
(74)
) over Gaussian noise whose width is related to h. Since by the virial
where E = l/u,
(75)
+ gx3 = rlnr
which allows a solution by strong coupling expansion for large g. The result is
c>
l/3
xn=
rln
g
-(r/;!3,77,-*/3
3g2/3
q,'J3)
(76)
0(E2>.
expectation
> are
(77)
On the lattice the path integral becomes a product of ordinary integrals which can be performed
287
with:
(78)
The ground state energy on the lattice regulated theory then has the form
E. = Jsz
-&nz,
(79)
n=o
where
is a dimensionless length. The critical index can be determined from the logarithmic derivative of E,,
with respect to z. Using PadC approximants to extrapolate the lattice series to small lattice spacing
we found that [ 211
E. =
ca.16
(80)
verifying that SUSY is unbroken in the continuum limit. Using the same methods for the case
W(x) = 8x2/2 we were able to verify that the ground state energy was not zero as we took the
continuum limit. After verifying the applicability of this method in SUSY QM, Bender et al. then
successfully used this method to study non-perturbative SUSY breaking in Wess-Zumino models of
field theory [ 131.
3. Factorization
In the previous section we found that once we know the ground state wave function corresponding
to a Hamiltonian Hi we can find the superpotential Wr (x) from Eq. (6). The resulting operators Al
and Al obtained from Eq. (4) can be used to factorize Hamiltonian Hr. We also know that the ground
state wave function of the partner Hamiltonian H2 is determined from the first excited state of HI via
the application of the operator A,. This allows a refactorization of the second Hamiltonian in terms of
IV,. The partner of this refactorization is now another Hamiltonian H3. Each of the new Hamiltonians
has one fewer bound state, so that this process can be continued until the number of bound states
is exhausted. Thus if one has an exactly solvable potential problem for HI, one can solve for the
energy eigenvalues and wave functions for the entire hierarchy of Hamiltonians created by repeated
refactorizations.
Conversely if we know the ground state wave functions for all the Hamiltonians in
this hierarchy, we can reconstruct the solutions of the original problem. Let us now be more specific.
From the last section we have seen that if the ground state energy of a Hamiltonian HI is zero then
it can always be written in a factorizable form as a product of a pair of linear differential operators.
It is then clear that if the ground state energy of a Hamiltonian HI is Ei with eigenfunction +i
then in view of Eq. (3), it can always be written in the form (unless stated otherwise, from now on
we set h = 2m = 1 for simplicity):
H, = AtAl + E;, = -$+w,
(81)
288
where
AI = $
+W,(x),
A;=--$+W,(x),
WI(X) =-
d In$~
&
(82)
&=A,Ai+E,,
(83)
where
V*(x)
= Wf + W: + I$
= V,(x)
2W;= Vj(x) - 2$
In$i.
(84)
We will introduce the notation that in Ei), rz denotes the energy level and (m) refers to the mth
Hamiltonian H,. From Sec. 2, the energy eigenvalues and eigenfunctions of the two Hamiltonians
H, and Hz are related by
,$:, = E;*,
(85)
Now starting from HZ whose ground state energy is Ei2 = E, one can similarly generate a third
Hamiltonian H3 as a SUSY partner of HZ since we can write HZ in the form:
H2=4A,+E0
I = A;A2 +
El,
(86)
where
A2
A; = -$
=$+W,(x),
+ W*(x),
W*(x)=-
d In $;*I
dx
(87)
(88)
where
v,(x)
w; + W;
El = vz(x) - 2-$Wi*)
= V,(X) - 2-$ln($$@$*)),
(89)
Furthermore
E3 = EC21 -- ,I$)
llf29
#i3, = g,
= (E$
Ei2) -r/*A2$;;;
_ El) +(
(90)
p (2
E(l), and eigenfunctions qQ:rwith 0 5 IE5 (p - I), then we can always generate a hierarchy of
(i - 1) Hamiltonians H2, . . . , HP such that the mth member of the hierarchy of Hamiltonians (H,,,)
F: Cooper et aLlPhysics
has the same eigenvalue spectrum as Ht except that the first (m - 1) eigenvalues
in H,. In particular, we can always write (m = 2,3, . . . , p ) :
H, = A;A,
+E;;!,
= -$
+ Vm(x),
289
of HI are missing
(91)
where
4, =
$ +wm(x),
Wm(x>= -
d In+!@)
dx
(92)
= ,@+I)
ntl
= . . . = E7;;,_1,
$()
= (E;;),_, - Ec12)-f2.. . (E;;,_, - E$*))-/2A,_, . . .A&,:,_,
n
V,(x)
= V,(x)
- 2$ln(&t
. . .&-1).
(93)
In this way, knowing all the eigenvalues and eigenfunctions of HI we immediately know all the
energy eigenvalues and eigenfunctions of the hierarchy of p - 1 Hamiltonians. Further the reflection
and transmission coefficients (or phase shifts) for the hierarchy of Hamiltonians can be obtained in
terms of RI, Tl of the first Hamiltonian HI by a repeated use of Eq. (30). In particular we find
(94)
where k and k are given by
k = [E-
(W)2]1/2,
4. Shape invariance
(95)
Most text books on quantum mechanics describe how the one dimensional harmonic oscillator
problem can be elegantly solved using the raising and lowering operator method. Using the ideas of
SUSY QM developed in Section 2 and an integrability condition called the shape invariance condition
[44], we now show that the operator method for the harmonic oscillator can be generalized to the
whole class of shape invariant potentials (SIP) which include all the popular, analytically solvable
potentials. Indeed, we shall see that for such potentials, the generalized operator method quickly yields
all the bound state energy eigenvalues, eigenfunctions as well as the scattering matrix. It turns out
that this approach is essentially equivalent to Schrodingers method of factorization [49,14] although
the language of SUSY is more appealing.
Let us now explain precisely what one means by shape invariance. If the pair of SUSY partner
potentials &,2(x) defined in Section 2 are similar in shape and differ only in the parameters that
290
appear in them, then they are said to be shape invariant. More precisely,
V,,, ( X; al ) satisfy the condition
V2(KUl) =
K(xa2)
+Na,),
(96)
= 0,
&)(~;a~)
= Nexp
[-
jWi(%ui)&]
.
and eigenfunctions
(97)
We now show that the entire spectrum of Hi can be very easily obtained algebraically by using
the shape invariance condition (96). To that purpose, let us construct a series of Hamiltonians H,,
In particular, following the discussion of the last section it is clear that if HI has p
s = 1,2,3,...
bound states then one can construct p such Hamiltonians HI, H2, . . . , HP and the nth Hamiltonian
H, will have the same spectrum as HI except that the first n - 1 levels of HI will be absent in H,.
On repeatedly using the shape invariance condition (96)) it is then clear that
s-l
H$=-$
+ Vl(x;u,)
+ CR(Uk),
(98)
kl
where a, = f- (al ) i.e. the function f applied s - 1 times. Let us compare the spectrum of H, and
H s+l. In view of Eqs. (96) and (98) we have
H s+l =-$+v,(x;us+,)
+&(uk)
k=I
-$
V,(x;a,)
(99)
gR(ak).
k=l
(100)
k=l
This follows from Eq. (98) and the fact that E,(I - 0 . On going back from H, to H,_, etc, we would
eventually reach H2 and HI whose ground state energy is zero and whose nth level is coincident
291
with the ground state of the Hamiltonian H,,. Hence the complete eigenvalue spectrum of Hi is given
bY
E,;(~I) = kR(n,);
E,-(a,) = 0.
(101)
k=l
We now show that, similar to the case of the one dimensional harmonic oscillator, the bound state
wave functions (cln(l)
(x; al) for any shape invariant potential can also be easily obtained from its
ground state wave function +$ (x; ai) which in turn is known in terms of the superpotential. This
is possible because the operators A and A+ link up the eigenfunctions of the same energy for the
SUSY partner Hamiltonians H i,*. Let us start from the Hamiltonian H, as given by Eq. (98) whose
ground state eigenfunction is then given by @i(x; a,). On going from H, to Hs_I to Hz to HI and
using Eq. (14) we then find that the nth state unnormalized, energy eigenfunction $A)(x; al) for
the original Hamiltonian HI (x; al) is given by
fi;l)(x;a,)
0; A+(x;a,)A+(x;a2)
~~~A+(x;u,)~~~(x;u,+~),
(102)
which is clearly a generalization of the operator method of constructing the energy eigenfunctions for
the one dimensional harmonic oscillator.
It is often convenient to have explicit expressions for the wave functions. In that case, instead of
using the above equation, it is far simpler to use the identify [46]
t,b;(x;ul) =A+(x;u~)@;!?,(wd.
(103)
Finally, it is worth noting that in view of the shape invariance condition (96), the relation (30)
between scattering amplitudes takes a particularly simple form
R, (k ~1) =
T,(k;a,)
W_(Q)
+ ik
W_(Q)
- ik >
W+(al)
- ik
W_(a,)
- ik >
Rl(k;ud,
(105)
T,(k;ud,
thereby relating the reflection and transmission coefficients of the same Hamiltonian HI at al and
a2(=
f(Q>>.
+R(Q)
(106)
or equivalently
W2(x;al)
+ W(x;u,)
= l?2(~;u,)
- @(~;a,)
+ R(q).
(107)
292
K(x;a2)
(108)
+Rad,
that is
tV(x;
(109)
We now show that when this condition holds, the energy eigenvalues and eigenfunctions of the
potential V, (x; aI ) can be obtained algebraically. First of all, let us notice that unbroken SUSY implies
zero energy ground states for the potentials V,(x; aI) and fi (x; al) :
Ep(q)
= 0,
The degeneracy
Ep(a,)
(110)
E(2)(u~)
= @,(a,);
n
From Eq. (106)
IP2(uJ
n
= 0.
E;2(al) = E$J&l,).
it follows that
= B(U,)
n
(112)
+ R(u,) 3
E[)(uJ
yield
= I?($).
(113)
(111)
= E(U2)
n
(108) yields
+ I?(&).
(114)
= E;l,(a:!)
+&a,)
(11%
+ Ru,).
we obtain (n = 0, 1,2, . .
J
(116)
W&t,l).
(117)
We now show that, similar to the discussion of the last subsection, the bound state wave functions
$() (x; al) can also be easily obtained in terms of the ground state wave functions I,@)(x; al) and
&-?r, (x; aI ) which in turn are known in terms of the superpotentials W and w. In particular from Eq.
(106) it follows that
rC/;;;(x;ur)
0: A+(x;~,)t,b~~(x;u~)
0: A+(x;u,)~~)(x;u,),
(118)
0; ij+(x;u,)$~(x;u~).
(119)
cc ij+(x;u,)@)(X;u~)
293
(120)
t,hll()(x;
a,)
0:
A+(x; a,)&)(~;
al).
(123)
Finally, it is easily shown that the relation (30) between the scattering amplitudes takes a particularly simple form
Rl(kh)
T,(k;ul)
W_(Q)
+ ik
Iv-(al)
+ ik
W_(u*)
I%-(q)
ik
ik
W+(u,)
- ik
@+(a,) - ik
W_(q)
- ik
l?_(q)
- ik
R1(ka2),
(124)
Tl(k;uz),
(125)
thereby relating the reflection and transmission coefficients of the same Hamiltonian at ai and u2.
It is clear that this procedure can be easily generalized and one can consider multi-step shape
invariant potentials and in these cases too the spectrum, the eigenfunctions and the scattering matrix
can be obtained algebraically.
4.3. Strategies for categorizing shape invariant potentials
Let us now discuss the interesting question of the classification of various solutions to the shape
invariance condition (96). This is clearly an important problem because once such a classification
is available, then one discovers new SIPS which are solvable by purely algebraic methods. Although
the general problem is still unsolved, two classes of solutions have been found and discussed. In the
first class, the parameters al and u2 are related to each other by translation (a2 = al + cr) [52,53].
Remarkably enough, all well known analytically solvable potentials found in most text books on
nonrelativistic quantum mechanics belong to this class. Last year, a second class of solutions was
discovered in which the parameters al and u2 are related by scaling (~22= +zl) [ 58,591.
4.3. I. Solutions
involving translation
We shall now point out the key steps that go into the classification of SIPS in case u2 = al + LY
[52]. Firstly one notices the fact that the eigenvalue spectrum of the Schrodinger equation is always
such that the nth eigenvalue E,, for large n obeys the constraint [ 1331
l/n2 5 E,, 5 n2,
(126)
294
where the upper bound is saturated by the square well potential and the lower bound is saturated by
the Coulomb potential. Thus, for any SIP, the structure of E,, for large IZis expected to be of the form
En-
c Cnna,
a
-25a52.
(127)
Now, since for any SIP, E, is given by Eq. (lOl), it follows that if
R(G)
c kY
(128)
then
-35y51.
(129)
How does one implement this constraint on R(ak)? While one has no rigorous answer to this
question, it is easily seen that a fairly general factorizable form of W(x; al) which produces the
above k-dependence in R(ak) is given by
W(x; al 1 =
fi(x)
(130)
i=l
where
al =
(131)
with ci, a, j? being constants. Note that this ansatz excludes all potentials leading to En which contain
fractional powers of n. On using the above ansatz for W in the shape invariance condition (96)
one can obtain the conditions to be satisfied by the functions gi( x) , hi(x) , fi( X) e One important
condition is of course that only those superpotentials W are admissible which give a square integrable
ground state wave function. It turns out that there are no solutions in case m 2 3 in Eq. (130), while
there are only two solutions in case m = 2 i.e. when
W-w)
(132)
B) = Atanhar
- Bcothcwr,
A > Z3> 0,
(133)
and
W(x;A,B)
=Atanax-Bcotax;
A,B>O,
(134)
where 0 < x < 7r/2a and $(x = 0) = $(x = 7r/2a) = 0. For the simplest possibility of m = 1, one
has a number of solutions to the shape invariance condition (96). In Table 4.1, we give expressions
for the various shape invariant potentials V,(x) , superpotentials W(x) , parameters al and a2 and the
corresponding energy eigenvalues EL) [ 54,551.
As an illustration, let us consider the superpotential given in Eq. ( 134). The corresponding partner
potentials are
v,tx;A, B)
= -(A
VZ(X; A, B) = -(A
=Vi(x;A+a,B+a)
295
(135)
+(A+B+2a)*-
(A+B)*
(136)
(137)
the bound state energy eigenvalues of the potential V, (x; A, B) are then given
bY
E; = =&(a,)
= (A + B + 2ncr)* - (A + B)*.
(138)
k=l
B) cx (cosax)S(sincux)A
W as given by
(139)
where
s = A/a;
h = B/CL
(140)
The requirement of A, B > 0 that we have assumed in Eq. (134) guarantees that & (x; A, B) is
well behaved and hence acceptable as x -+ 0,~/2a.
Using this expression for the ground state wave
function and Eq. ( 103) one can also obtain explicit expressions for the bound state eigenfunctions
@A)(x; A, B). In particular, in this case, Eq. (103) takes the form
&II--C {a&
= (-2
+ Atanax
- Bcotax
+4,+,(x; {a*}).
>
(141)
(142)
h))
we obtain
= a( 1 - y) dR,_1 (y; A + a, B + a)
dy
B) - (a+B+cu)ylR,-l(y;A+cr,B+cu).
B) = (1 - y)/*(l
(143)
+y)2P,-/2,-/2(y).
(144)
so that the unnor-
(145)
296
Table 4.1
All known shape invariant potentials in which the parameters u2 and al are related by a translation (~2 = al + a). The
energy eigenvalues and eigenfunctions are given in units fi = 2m = 1. The constants A, B, (Y,OJ,1 are all taken 2 0. Unless
otherwise stated, the range of potentials is --co 5 x 5 00, 0 5 r 5 CO.For spherically symmetric potentials, the full wave
function is &,lrn( r, 0,4) = cCr,l(
r) &( 0, 4). Note that the wave functions for the first four potentials (Hermite and Laguerre
Name of potential
Potential VI(x; al )
shifted oscillator
+x-b
fW
three dimensional
itir------(I+ 1)
r
+w2r2 + v
al
012
- (1+3/2)0
0
1
oscillator
cl+11
&l(-tl)
rf+7
e4
____
Coulomb
--- e2
2(Z+ 1)
Morse
A - Bexp(-ax)
A2 + B2 exp( -2ax)
x exp( -ax)
Scarf II
(hyperbolic)
A2 + ( BZ - A* - Aa)sech2ax
+B (2A + a) sech LYX
tanh (YX
Rosen-Morse II
(hyperbolic)
A tanh LYX
+ B/A
(B < A)
A2 f B*jA2 - A( A + a)sech2cYx
+ 2B tanh (YX
Eckart
-A coth cu + B/A
(B > A2)
A2 + B2/A2 + A( A - (~)cosech~crr
- 2B coth ar
Scarf I
(trigonometric)
generalized PiischlTeller
A2 + ( B2 + A2 + Aa)cosect?ar
- B (2A + a) coth ar cosech ar
Rosen-Morse I
(trigonometric)
-Acotax - B/A
(0 < (Yx< 7r)
+ 4(1+ 1>2
- 2B(A + a/2)
The procedure outlined above has been applied to all known SIPS [46,125] and the energy eigenfunctions
$j) (y) have been obtained in Table 4.1, where we also give the variable y for each
case.
Several remarks are in order at this time.
(i) The PGschl-Teller I and II superpotentials as given by Eqs. ( 134) and ( 133) respectively have
not been included in Table 4.1 since they are equivalent to the Scarf I (trigonometric)
and
generalized Pijschl-Teller superpotentials
W, = -Atanax+Bsecax,
297
polynomials)
are special cases of the confluent hypergeometric function while the rest (Jacobi polynomials)
are special
cases of the hypergeometric function. Fig. 5.1, taken from Ref. [ 1291, shows the inter-relations between all the SIPS in the
table via point canonical coordinate transformations.
a2
Eigenvalue
nw
If1
I?,$
Variable y
y=
(+))/2
x-
I?!
0
>
y = +r2
2nw
e4
--
1+1
;T
( (Ifl)Z
(n+l+
exp(-$*)Hdy)
1)2 >
y=
y+exp(-$y)Li+(y)
(n+Y+l)
A-a
A2 - (A -n(u)
y = ( ~B/LY)exp( --ax),
s=A/a
YS-Rexp( -$y)LF-*(y)
A-a
A2 - (A - na)*
y = sinh (YX,
s = A/a, A = B/a
y = tanhcyn,
( 1 _ y) (s-n+0)/2 ( 1 + y) (s-n-n)/2
x pb-Rfw-n-a)
Cy)
n
A-a
A2 - (A - n(u) + B2/A2
- B/(A
s = Ala,
- TUY)~
A = B/cx2,
a = A/(s - n)
AS-a
Afa
A2 - (A + n(u) + B2/A2
- B*/( A + na)*
y = coth Lyr,
s = A/a, A = B/a,
a = A/(n + s)
ty _
(A + na) - A*
y = sin ffx,
( 1 _ y) (s-A)/2 ( 1 + y) (s+A)/z
x p~s-A-1/2,s+A-l/2) cy)
s=A/a,
A-a
A+a
A2 - (A -n(u)
(A + n(u) - A2 + B2/A2
_ B2/(A + FKX)~
A=B/a
l)-(s+n-n)/2(Y
redefinition
l)-_(s+n+a)/*
(Y)
y = cash ur,
s=AIa, A=B/a
(y _ 1)b-~/2(~ + l)-(+s)/2
x P(A--s--1/2.-A--s--l/2) (y)
y = icot ffx
s=A/cu, A=B/a,
a=A/(s+n)
x P~-S-+-S--R)
of the parameters
(146)
[ 951. For example, one can write
(147)
(ii)
which is just the PGschl-Teller II superpotential of Eq. (133) with redefined parameters.
Throughout this section we have used the convention of ti = 2m = 1. It would naively appear
that if we had not put ti = 1, then the shape invariant potentials as given in Table 4.1 would
E Cooper
298
all be h-dependent. However, it is worth noting th?t in each and every case, the h-dependence
is only in the constant multiplying the x-dependent function so that in each case we can
always redefine the constant multiplying the function and obtain an h-independent potential.
For example, corresponding to the superpotential given by Eq. ( 134)) the h-dependent potential
is given by (2m = 1)
V,(x;A,B)
=W2-hW
+ B ( B + ha) cosec2ax.
(148)
On redefining
A(A+hia)
=a;
B(B+tia)
=b,
(149)
W = n tanhx,
(150)
where FZis positive integer (1,2,3,. . .). The two partner potentials
Vi(x; n) = n2 - n(n + l)sech2x,
(151)
n,
u2=n-
1.
(152)
299
On going from VI to V, to V, etc., we will finally reach the free particle potential which is reflectionless
and for which T = 1. Thus we immediately conclude that the series of potentials V, V,, . . . are all
reflectionless and the transmission coefficient of the reflectionless potential V,(x; n) is given by
T,(k
=
n>= (-n-ik)(-n+l
r(-n
- ik)r(n+
r( -ik)T(
-ik)...(-1
-ik)
1 - ik)
1 - ik)
(153)
The scattering amplitudes of the Coulomb [47] and the potential corresponding to W = A tanhx +
B sech x [ 481 have also been obtained in this way.
There is, however, a straightforward method [48] for calculating the scattering amplitudes by
making use of the nth state wave functions as given in Table 4.1. In order to impose boundary
conditions appropriate to the scattering problem, two modifications of the bound state wave functions
have to be made: (i) instead of the parameter y1 labelling the number of nodes, one must use the
wave number k so that the asymptotic behaviour is exp(ikx) as x --+ 00. (ii) the second solution
of the Schrodinger equation must be kept (it had been discarded for bound state problems since it
diverged asymptotically).
In this way the scattering amplitude of all the SIPS of Table 4.1 have been
calculated in Ref. [48].
4.3.2. Solutions involving scaling
For almost nine years, it was believed that the only shape invariant potentials are those given
in Table 4.1 and that there were no more shape invariant potentials. However, very recently we
have been able to discover a huge class of new shape invariant potentials [ 58,591. It turns out that
for many of these new shape invariant potentials, the parameters a2 and al are related by scaling
( a2 = qal, 0 < q < 1) rather than by translation, a choice motivated by the recent interest in qdeformed Lie algebras. We shall see that many of these potentials are reflectionless and have an
infinite number of bound states. So far none of these potentials have been obtained in a closed form
but are obtained only in a series form.
Let us consider an expansion of the super-potential of the from
W(x; a,) = ggj(x)a{
(154)
j=O
94,
O<q<l.
(155)
(156)
and equating powers of al yields [59,58]
2&(x)
= R,;
g\(x)
= r14,
(157)
drtCgi(x)gn-j(x)9
(158)
24go(x)g1(xI
n-l
g:(X)
2&gO(x)gn(x)
= Ynd~ -
j=I
where
r, z &/(I
-qn),
d,=(l-g)/(l+q),
n=1,2,3
,...
(159)
This set of linear differential equations is easily solvable in succession to give a general solution of
Eq. (96). Let us first consider the special case gO(x) = 0, which corresponds to Ra = 0. The general
solution of Eq. (158) then turns out to be
n-l
g,(x) = 4
Y, - Cgj(x)gn-i(X)
j=l
IZ= 172,
a a.
(160)
where without loss of generality we have assumed the constants of integration to be zero. We thus see
that once a set of Y, are chosen, then the shape invariance condition essentially fixes the g,(x) (and
hence W( x; al ) ) and determines the shape invariant potential. Implicit constraints on this choice are
that the resulting ground state wave function be normalizable and the spectrum be sensibly ordered
which is ensured if R(qa,)
> 0.
The simplest case is r-1 > 0 and r, = 0, n > 2. In this case the Eq. (160) takes a particularly
simple form
g, (xl
= PnX2n-1 7
(161)
where
(162)
/3n= -
PI = d,rl,
and hence
W(x;a,)
= Ffija{x-
= &F(&x).
(163)
j=I
&W<g$, a),
t 164)
of Shabat and Spiridonov [ 60,611. It is worth pointing out
shown to satisfy q-supersymmetry
[62]. It may be noted
>_ 2, if any one r, (say rj) is taken to be nonzero then
[59,58] and in these instances the results obtained from
301
shape invariance and self-similarity are entirely equivalent and the Shabat-Spiridonov self-similarity
condition turns out to be a special case of the shape invariance condition.
It must be emphasized here that shape invariance is a much more general concept than selfsimilarity. For example, if we choose more than one Y, to be nonzero, then SIP are obtained which
are not contained within the self-similar ansatz. Consider for example, r, = 0, y1> 3. Using Eq. ( 160)
one can readily calculate all the g,(x), of which the first three are
!?I(x) =
g2(x)
4r1-G
g3(x) = -;d,r,d2r2d3x3
d2r,x - ;d$-;x3,
+ $d;r;d2d3x5.
(165)
Notice that in this case W(x) contains only odd powers of x. This makes the potentials &,2(x)
symmetric in x and also guarantees unbroken SUSY. The energy eigenvalues follow immediately
from Eqs. (101) and (156) and are given by (0 < 9 < 1)
E()(ul)
(1 -t4)(1-4)
(l-4)
=r,
+r
(1+9%1 -s2>
(l-s2)
n=0,1,2,...
(166)
where rl = dl rlal
f2 = d2r2ui
-~(I.~+~2)+~(dzT:+2d~~~~2+4r:)+0(x6)
1.
(167)
The wave functions for the excited states can be recursively calculated from the relation ( 103).
We can also calculate the transmission coefficient of this symmetric potential (k = k) by using
the relation ( 125) and the fact that for this SIP a2 = qal . Repeated application of the relation ( 125)
gives
where
W(m,
Uj)
= 4s.
(169)
Now, as n --) co, u,,+~ = qnul --+ O(0 < q < 1) and, since we have taken go(x) = 0, one gets
W(x; u,,+~) --f 0. This corresponds to a free particle for which the reflection coefficient RI (k; al)
vanishes and the transmission coefficient is given by
7,(k;al)
O [ik- W(m,uj)]
=n
j=l [ik+W(myuj)]
(170)
The above discussion keeping only F-~,r2 j 0 can be readily generalized to an arbitrary number of
nonzero rj. The energy eigenvalues for this case are given by (rj z djrj&i)
Ei)(U*)
C
j
(1+4>(1- $7
rj
(I--@)
n=0,1,2,...
(171)
302
All these potentials are also symmetric and reflectionless with Tr as given by Eq. ( 170). The limits
and q -+ 1 of all these potentials are simple and quite interesting. At q = 1, the solution
of the shape invariance condition (96) is the standard one dimensional harmonic oscillator with
W(x) = Rx/2 while in the limit q -+ 0 the solution is the Rosen-Morse superpotential corresponding
to the one soliton solution given by
q ----t0
W(x) = fitanh(fix).
(172)
Hence the general solution as obtained above with 0 < q < 1 can be regarded as the multi-parameter
deformation of the hyperbolic tangent function with q acting as the deformation parameter. It is also
worth noting that the number of bound states increase discontinuously from just one at q = 0 to
infinity for q > 0. Further, whereas for q = 1 the spectrum is purely discrete, for q even slightly less
than one, we have the discrete as well as the continuous spectra.
Finally, let us consider the solution to the shape invariance condition (96) in the case when R,, $0.
From Eq. ( 157) it then follows that go(x) = Rex/2 rather than being zero. One can again solve the
set of linear differential equations ( 157) and ( 158) in succession yielding g1 (x) , g2( x) , . . . Further,
the spectrum can be immediately obtained by using Eqs. (101) and (156). For example, in the case
of an arbitrary number of nonzero Rj (in addition to Ro), it is given by
(173)
which is the spectrum of a q-deformed harmonic oscillator [ 136,137]. It is worth pointing out here
that, unlike the usual q-oscillator where the space is noncommutative
but the potential is normal
(w2x2), in our approach the space is commutative, but the potential is deformed, giving rise to a
multi-parameter deformed oscillator spectrum.
An unfortunate feature of the new SIPS obtained above is that they are not explicitly known in terms
of elementary functions but only as a Taylor series about x = 0. Questions about series convergence
naturally arise. Numerical solutions pose no serious problems. As a consistency check, Barclay et al.
have checked numerically that the Schrodinger equation solved with numerically obtained potentials
indeed has the analytical energy eigenvalues given above. From numerical calculations one finds that
the superpotential and the potential are as shown in Figs. 4.1 and 4.2 [ 591 corresponding to the case
when rl $0, r, = 0, n > 2. These authors, however, see no evidence of the oscillations in W and V as
reported by Degasperis and Shabat [ 631. A very unusual new shape invariant potential has also been
obtained [59] corresponding to r1 = 1, r2 = -1, r, = 0, n > 3 (with q = 0.3 and a = 0.75) which is
shown in Fig. 4.3. In this case, whereas Vi (x) is a double well potential, its shape invariant partner
potential V2( x) is a single well.
It is worth pointing out that even though the potentials are not known in a closed form in terms
of elementary functions, the fact that these are reflectionless symmetric potentials can be used to
constrain them quite strongly. This is because, if we regard them as a solution of the KdV equation
at time t = 0, then being reflectionless, it is well known that such solutions as t --f &co will break up
into an infinite number of solitons of the form 2k?sech*kix [ 138,139]. On using the fact that the KdV
solitons obey an infinite number of conservation laws corresponding to mass, momentum, energy ....
one can immediately obtain constraints on the reflectionless SIPS obtained above [ 591.
303
2.0
1.5
W
1.0
0.5
0.0
about x = 0) corresponding
to the superpotentials
0.1
0.0
-0.1
-30
-20
-10
10
20
30
Fig. 4.3. A double well potential V,(x) (solid line) and its single well supersymmetric partner h(x)
(dotted line). Note
that these two potentials are shape invariant [59] with a scaling change of parameters. The energy levels of VI (x) are
clearly marked.
304
W(X;Ul)
W(X;Ul) =
&j(X)U{;
j=O
(174)
Fhj(X)Uj
j=O
=gRjd;
.i=o
.R(u~) =gRjd
(175)
j=O
Using these in Eqs. (107) and (109) and equating powers of al yields (n=0,1,2,. . .)
gk + f:
gjg,-j
= 2
j=O
hk + 2
hjh,-j
- hi + R,,
(176)
j=O
hjh,_j = q 2
(177)
j=O
j=O
This set of linear equations is easily solved in succession. For example when R. = & = 0 (and hence
go(x) = ho(x) = 0) and further R, = I?, = 0, n 2 3 one can readily calculate all the g,(x) and h,(x) .
The first two of each are
(RI-&)
gl=
(1-q)
x9
CR2- &)
g2=
(1_q2)
x3
+ 3(1 - q)3
h,=(hGG)
I(1 - q)(&
- R:) -2(l
+q)R#d,
(178)
(1 -s>
x7
R2x
h*=
(1
-
-q2)
X3
3(1 +q)(l
-q)*
[(l+q)&+(l+q)(l-q2)R;-2q(l-q)Rlk,].
(179)
The energy eigenvalues which follow from Eqs. ( 117) and ( 118) are [ 591
(180)
For the special case of R2 = R2 = 0 the spectrum was obtained previously by Spiridonov from the
considerations of the two-step self-similar potentials [ 1351.
305
potential
(181)
provided R = 31?. This procedure can be immediately generalized and one can consider shape
invariance with a scaling ansatz in 3,4, . . . , p steps and thereby obtain multi-parameter deformations
of the 3,4,. . . , p soliton Rosen-Morse potential.
4.3.4. Other solutions
So far we have obtained solutions where a2 and al are related either by scaling or by translation.
Are there shape invariant potential where a2 and al are neither related by scaling nor by translation?
It turns out that there are other possibilities for obtaining new shape invariant potentials. Some of the
other possibilities are: a2 = qay with p = 2,3, . . . and a2 = qaI /( 1 + pal). Let us first consider the
case when
a2 = qa:
(182)
2nt
&,(x>
Cgj(x)g2m-j(X)
= 4
~~j(x)grr-j(~)
fl&(~)
R2m7
(183)
j=O
j4
2m+ 1
&p,+,(x)
gj(x)g2m+l-j(X)
(184)
= R2m+l9
.i=o
which can be solved in succession and one can readily calculate all the g,(x) . For example,
only RI and R2 are nonzero, the first three gs are
gl (x) = Rlx,
when
(185)
= 0)
11=1,2,...
(186)
The q -+ 0 limit of these equations again correspond to the Rosen-Morse potential corresponding to
the one soliton solution. One can also consider shape invariance in multi-steps along with this ansatz
thereby obtaining deformations of the multi-soliton Rosen-Morse potential.
One can similarly consider solutions to the shape invariance condition (96) in case
9ar
al = ___
1 +pal
(187)
when 0 < q < 1 and pal < 1 so that one can expand ( 1 + pal) - in powers of al. For example,
when only R1 and R2 are nonzero, then one can show that the first two nonzero g, are
306
E Cooper
gz(x) =
Rz+---
P@I
(1 +41
et al. /Physics
(1 -s>
x3
(1 +@(1 +q2) 3
>
x-
267-385
(188)
spectrum is (Eh = 0)
(189)
Generalization to the case when several Rj are nonzero as well as shape invariance in multi-steps
is straightforward.
We would like to close this subsection with several comments.
(9 Just as we have obtained q-deformations of the reflectionless Rosen-Morse and harmonic oscillator potentials, can one also obtain deformations of the other SIPS given in Table 4.1?
(ii) Have we exhausted the list of SIP? We now have a significantly expanded list but it is clear
that the possibilities are far from exhausted. In fact it appears that there are an unusually
large number of shape invariant potentials, for all of which the whole spectrum can be obtained
algebraically. How does one classify all these potentials? Do these potentials include all solvable
potentials [ 52,53]?
(iii) For those SIP where a2 and al are not related by translation, the spectrum has so far only been
obtained algebrically. Can one solve the Schrodinger equation for these potentials directly?
There
is a fundamental difference between those shape invariant potentials for which a2 and
(iv)
al are related by translation and other choices (like a2 = qal). In particular, whereas in the
former case the potentials are explicitly known in a closed form in terms of simple functions,
in the other cases they are only known formally as a Taylor series. Secondly, whereas in the
latter case, all the SIP obtained so far have infinite number of bound states and are either
reflectionless (or have no scattering), in the former case one has also many SIP with nonzero
reflection coefficients.
4.4. Shape invariance and noncentral solvable potentials
We have seen that using the ideas of SUSY and shape invariance, a number of potential problems
can be solved algebraically. Most of these potentials are either one dimensional or are central potentials
which are again essentially one dimensional but on the half line. It may be worthwhile to enquire
if one can also algebraically solve some noncentral but separable potential problems. As has been
shown recently [ 1311, the answer to the question is yes. It turns out that the problem is algebraically
solvable so long as the separated problems for each of the coordinates belong to the class of SIP. As
an illustration, let us discuss noncentral separable potentials in spherical polar coordinates.
In spherical polar coordinates the most general potential for which the Schrodinger equation is
separable is given by [ 1401
v(e)
--$-
V(4)
+
r2
(190)
307
where V(r),V(8)
and V(4) are arbitrary functions of their argument. First, let us see why the
Schrijdinger equation with a potential of the form given by Eq. (190) is separable in the (r, 13,4)
coordinates. The equation for the wave function $(r, B,4) is
(191)
It is convenient
to write +( Y,8, 4) as
(192)
Substituting Eq. ( 192) in Eq. ( 191) and using the standard separation of variables procedure,
obtains the following equations for the functions K(4), H(B) and R(r):
-g
-g
one
V(4)K(4J) = m2w$>
(193)
(194)
-2 + V(r)
[
(12- )
r2
R(r) = H?(r),
(195)
308
The important point to note is that three particles in one dimension, after the center of mass
motion is eliminated, have two degrees of freedom. This may therefore be mapped on to a one-body
problem in two dimensions. Calogero [ 1411 considered the case where the two dimensional potential
is noncentral but separable in polar coordinates r, 4. From the above discussion, it is clear that if
the potentials in each of the coordinates r and 4 are chosen to be shape invariant, then the whole
problem can be solved exactly.
Calogeros [ 1411 solution of the three-body problem is for the potential
Vc =w~/SC(X~-X~)~+~C(X~-X~)-~,
i<j
(196)
i<j
to avoid a collapse of the system. Wolfes [ 1421 showed that a three-body potential
v,=f[(X1+X2-2X3)-2+(X~+Xg-2X1)-2+(X3+X,-2X2)-21
(197)
is also solvable when it is added to V,, with or without the pairwise centrifugal term. The last
two terms on the right-hand side of Eq. (197) are just cyclic permutations of the first. Henceforth,
such terms occurring in any potential are referred to as cyclic terms. In this subsection, we give
more examples of three-body potentials that can be solved exactly. Our first solvable example is the
three-body potential of the form
V,
af1
2r2
( + x2 - 2x3)+
(x1
cyclic
terms
x2)
19
(198)
which is added to the Calogero potential Vc of Eq. ( 196). In the above equation,
y2 = ir (Xi -
x212
(x2
x3j2
(x3
xd21.
(199)
TO see why potentials given by Eqs. (196) to (198) are solvable, define the Jacobi coordinates
R=;(xI+x~+x~),
x =
(xl
~21,
(200)
y =
(x1
Jz
+X2
&
2x3)
(201)
*
After elimination of the center of mass part from the Hamiltonian, only the x- and y-degrees of
freedom remain, which may be mapped into polar coordinates
x = rsin4,
y = rcos+.
(202)
Obviously, the variables r, 4 have ranges 0 _<r 5 00 and 0 5 4 5 27r. It is straightforward to show
that
(xi -x2)
= JZrsin4,
JZrsin(4
+ 47r/3).
(203)
It turns out that Vc, VW as well as & are all noncentral but separable potentials in the polar coordinates
r,$. As a result, the Schrodinger equation separates cleanly in the radial and angular variables, and
the wave function can be written as
309
(204)
In all three cases, the radial wave function obeys the equation
-G
d*
3
+ ,wr+
(B: - $)
r*
&I = G&l(r)
(205)
where Bf- is the eigenvalue of the Schrijdinger equation in the angular variable. Eq. (205) corresponds
to a SIP and the eigenvalues E,,, and the eigenmnctions R,,, which follow from Table 4.1 are
En,=
Jl3 20(2n+B1+1),
n,Z=0,1,2
,...
R,, =rexp[-~(J3/2)wr*]L~[$(@)wr*],
(207)
where BI> 0.To examine the angular part of the eigenfunction Fl( 4), take the potential
Then, in the variable 4, the Schriidinger equation is
--$+g??I=1cosec* [@+2(m-
(206)
1):
V, + V,.
1);
(208)
(209)
[4+2(m-
+&cot
!?I=1
cosec2
[ q5 +
m=l
-&ot[&+2(m-
1)7r/3] =3cot3+J
(210)
tll=l
--$ +$cosec*3+
!fl
cot 34
F1= Bffi(4,).
(211)
Now this is again a SIP and hence its eigenvalues and eigenfunctions are
+ a + 1/2)2,
(212)
(213)
l/2( 1 + 2g)/*.
(214)
It must be noted that g > -l/2 for meaningful solutions. As expected, we recover the results for the
Calogero potential Vc in the limit fl = 0.
310
For distinguishable particles, a given value of 4 defines a specific ordering. For 0 5 4 5 7r/3, Eq.
(203) implies X1 2 X2 2 X3, and other ranges of 4 correspond to different orderings [ 141,142].
For singular repulsive potentials, crossing is not allowed, and Fl( 4) of Eq. (213) is zero outside
0 < 4 5 r/3. Following Calogero, the wave function for the other ranges may be constructed.
Similarly, for indistiguishable particles, symmetrized or antisymmetrized
wave functions may be
constructed.
Proceeding in the same way and using the results of Table 4.1, it is easily shown that the other
exactly solvable potentials are [ 1321
~=~W2C(X~-Xj)2+3gt(X*+X2-2X3)-2+cyclicterms]
i<j
--_
3&f,
2
V3 =
(Xl
i (x,
r2
$0 C(Xi
x2>
+ gC(Xi -
Xj)2
i<j
&r
V, =
io2
[
C(Xi
1,
(215)
Xj)-2
i<j
(Xl +-x2
f3
--
+ cyclic terms
+X2-2x3)
(x1 -
2x3)
+
cyclic terms
x2)2
(216)
,
1
i<j
(XI
x2)
(X, + X2 - 2x3)2
+ cyclic terms
1.
(217)
Notice that in all these cases one has combined the SIP in the angular variable with the harmonic
confinement. The three body scattering problem has also been studied in these cases after droping the
harmonic term [ 1321. The possibility of replacing the harmonic confinement term with the attractive
i-type interaction has also been considered. Note that in this case one has both discrete and continuous
spectra. Further, in this case one can obtain exact solutions of three-body problems for all the SIPS
discussed above (i.e. V,, . . . I& Vc, VW, Vc + VW) along with the attractive l/r potential [ 1321.
E Cooper et d/Physics
311
are only implicitly determined, while if the relationship is explicit then the newly found solvable
potentials are also shape invariant [ 127-1291. In a more specific special application of these ideas,
Kostelecky et al. [ 1301 were able to relate, using an explicit coordinate transformation, the Coulomb
problem in d dimensions with the d-dimensional harmonic oscillator. Other explicit applications of
the coordinate transformation idea are found in the review article of Haymaker and Rau [ 1201.
Let us see how this works. We start from the one-dimensional Schriidinger equation
-2 +[V(x)
f(z) =
=o.
- En1 &(x)
transformation
(218)
from x to z defined by
2,
(219)
so that
$=f-$
(220)
f d
d2
--___
dz2
f2 so
(221)
fdz+
To eliminate the first derivative term, one next rescales the wave function:
(222)
Adding a term E,,$ to both sides of the equation yields
(223)
where
V
%%) =f2-
f2
E,
4f2-2f
f"
1
l,f2
(224)
*
P(E,)
+ E, must be
(225)
is independent of n. How can one satisfy this condition? One way is to have f2 and G to have the
same functional dependence on x(z) as the original potential V. This further requires that in order
for P to be independent of n, the parameters of V must change with n so that the wave function
corresponding to the nth energy level of the new Hamiltonian is related to a wave function of the
old Hamiltonian with parameters which depend on ~1.This can be made clear by a simple example.
312
Let us consider an exactly solvable problem - the three dimensional harmonic oscillator in a given
angular momentum state with angular momentum p. The reduced ground state wave function for that
angular momentum is
tlrg(r) = yP+te-&2
W(r)
(226)
is given by
= ff?- - (p + 1)/r,
(227)
d2
PW+ 1)
HI=-%+
T-2
z = r2/2.
(230)
E, = 4an,
(231)
we want to satify is
V-4an+e,f2=G=s+Er2+R
Equating coefficients
D=P(P+
F = -4an
(232)
we obtain
l),
- 2a(P
E = E, + a2 = y,
+ 3/2) = -2Ze2.
(233)
We see that for the quantities D, E, F to be independent of n one needs to have that (Y, which
describes the strength of the oscillator potential, be dependent on n. Explicitly, solving the above
three equations and choosing fl= 21+ l/2, we obtain the relations:
Ze2
cu(zn)=2(l+n+l)
Z2e4
E=Y-4(E+l+n)2
(234)
We now choose y = ct? (1, n = 0) so that the ground state energy is zero. These energy levels are
those of the hydrogen atom. In fact, the new Hamiltonian written in terms of z is now
(235)
313
and the ground state wave function of the hydrogen atom is obtained from the ground state wave
function of the harmonic oscillator via
& =f
l/2flo
= Xl+le-nho)x~
(236)
Higher wave functions will have values of a which depend on IZso that the different wave functions
correspond to harmonic oscillator solutions with different strengths.
All the exactly solvable shape invariant potentials of Table 4.1 can be inter-related by point
canonical coordinate transformations [ 128,129]. This is nicely illustrated in Fig. 5.1. In general, r
cannot be explicitly found in terms of z, and one has
Jm,
dz/dr=f=
(237)
+ JAlog(r)
2
2
_ fi
log( 2 A + C r2 + 2&JA
+ Cr2 + Br4)
2
+ C log( C + 2Br2 + 2fi
JA + Cr2 + Br4)
40
(238)
This clearly is not invertible in general. If we choose this general coordinate transformation, then
the potential that one obtains is the particular class of Natanzon potentials whose wave functions are
confluent hypergeometric functions in the variable r and are thus only implicitly known in terms of
the true coordinate z. In fact, even the expression for the transformed potential is only known in
terms of r:
ii( z, D, E) = 1/f2 [ D/r2 + Er2 + F - f//4 + ff/2]
(239)
and thus only implicitly in terms of z. Equating coefficients, we get an implicit expression for the
eigenvalues:
CE,-F
2d_
- Jw=2n+1
(240)
as well as the state dependence on cx and p necessary for the new Hamiltonian to be energy
independent:
cy, = &7=-Z&,
p,=-1/2+JW.
(241)
The more general class of Natanzon potentials whose wave functions are hypergeometric functions
can be obtained by making an operator transformation of the generalized Poschl-Teller potential
whose Hamiltonian is:
(242)
314
P&hi-Teller-II
xc5
t
Generalized
Pikchl-Teller
Piischl-Teller-I
Scarf
:onometric)
(Trigonometric)
Rosen-Morse-II
Fig. 5.1. Diagram showing how all the shape invariant potentials of Table 4.1 are inter-related by point canonical coordinate
transformations. Potentials on the outer hexagon have eigenfunctions which are hypergeometric functions whereas those on
the inner triangle have eigenfunctions which are confluent hypergeometric functions. These two types are related by suitable
limiting procedures. The diagram is from Ref. [ 1291.
This corresponds
to a superpotential
(243)
W=atanhr-pcothr
and a ground state wave function given by:
& = sinhP YcoshPa r.
The energy eigenvalues
(244)
31.5
E, = (a - p)* - (a - ,B - 2n)*.
The most general transformation
equation is described by:
(245)
of coordinates
the Schrijdinger
fLB_
(246)
sinh* Y
From this we obtain an explicit expression
for z in terms of r.
-3A+B-C+(A+B-C)cosh2r
z = L&ul-l
J-2A
{ 24
+ 2B - C + 2( A + B) cash 2r + Ccosh* 2r
- filog{-A
+ 3B - C + (A + B + C) cosh2r
+ 2&+2A
+&log{A+B+Ccosh2r
+ fi
+ fi
-2A
log{2sinh*r}
(247)
&a-
1)
sinh* r
and thus only implicitly
eigenvalues:
[(y + l/2)*
- A#*
5.2. Generalizations
Y(Y-t
1)
cash* r
ff2
f"
+D-4f2+2f
- [ (6 - l/2)*
l/2,
on CYand /3 necessary
Bn = [ (6 - l/2)*
(248)
- BE,]~* + l/2.
for the
(249)
to be energy
(250)
In Ref. 1521 it was explicitly shown that the Ginocchio and Natanzon potentials, whose wave
functions are hypergeometric
functions of an implicitly determined variable are not shape invariant
and thus one could construct new exactly solvable potentials from these using the factorization
method. It is convenient to use the original approach [57,56] to define the Ginocchio and Natanzon
potentials and then obtain their generalization to more complicated solvable potentials whose wave
functions are sums of hypergeometric functions. These generalized potentials have coordinates which
are only implicitly known. The operator which allows one to construct the eigenfunctions
of H2
from those of HI converts single hypergeometric
functions of the implicitly known coordinate to
sums of hypergeometric
functions. This process yields totally new solvable potentials. The resulting
316
potentials are ratios of polynomials in the transformed coordinate in which the wave functions are
sums of hypergeometric functions. The transformed coordinate is only implicitly known in terms of
the coordinate system for the Schrodinger equation.
The original method of obtaining the Natanzon potentials was to find a coordinate transformation
that mapped the Schrijdinger equation onto the equation for the hypergeometric functions.
If we denote by Y the coordinate appearing in the Schrodinger equation:
[--$ivrrj]
lb(r> =o
(251)
describing
the hypergeometric
function
F(a,
/3; y; Z)
(252)
where 0 5 z 2 1, then the transformation
dz
dr=
22(1 -z)
R/2
of coordinates
is given by [ 561
R=az2+(c,-c~-a)z+co
- 1) + ho(1 -z)
uf(c1
-co>(2z
(253)
to a Schrodinger
+ h,z + 11/R
1)
z(z - 1)
(254)
where
We note that z is only implicitly known in terms of r. The potential V(r) is a function of six
dimensionless parameters f, ho, hi, a, COand cl. These six parameters will be related to the energy E,
and the parameters (a, p, 7) of the hypergeometric function below.
The class of potentials called the Ginocchio potentials is a subclass of the Natanzon class which
has only two independent variables nu and A where co = 0, cl = l/h4, a = cl - 1/A2, ho = -3/4,
hi = -1, and f= (v + 1/2)2 - 1.
The Ginocchio potentials are more easily discussed, however by considering them as derived from
coordinate transformations which map the Schrodinger equation into the differential equation for the
Gegenbauer polynomials.
The transformation is given by
(255)
$=(l-y2j[l+(A2-ljy2]
where now -1 5 y 2 1 and the potential is given by:
V(r)
= {-A*v(v
(256)
rA2 = tanh-(y)
- (1 - A2)/2tanhh1[(1
- A2)1/2y],
rA2 = tanh-(y)
+ (A2 - 1)12tan-[(A2
- 1)2y],
By introducing
g(y)
A < 1,
A > 1.
(257)
= 1 + (A2 - l)y2,
and changing the variable x to an angle x = cos 8, one obtains from the Schrodinger
equation for the Gegenbauer polynomials:
p2 $+(+P+lw2-
l/4
z1/2$,= 0
sin2e
317
equation,
the
(258)
where
z
I1 -
y2\12
= +2~4
A
The corresponding
(259)
energy eigenvalues
and eigenfuctions
are
~U,A~=[A~(~+1/2)~+(l-A~)(,+1/2)~]/~-(~+1/2),
(260)
(Gin= (1 -
(261)
Y2)~L./2[8(Y)~-(2~Lll~l)/4~,(~Ln+1/2)(X).
is:
;(l
(262)
- A2)y(y2 - 1) +poyA.
(254)
(263)
and the corresponding
(unnormalized)
energy eigenfunctions
- n; l+&;z_).
F(-n,a,
the superpotential
= [Soz - (1 +P,,>(l
- z)I/R~
(265)
- z>/(~R~~).
Once we have the superpotential and the explicit expression for the wave functions we can determine
all the wave functions of the partner potential using the operator:
Al =$+W(r)
=g$+w(z)
2z(l-2)
RI/2
d
z + W(z)
318
A,
dy d
+ W(y)
$ +W(r) = zdy
= (1 - y2>[1+
(A2 - l)y21-$
+ W(y)
for the case of the Ginocchio potentials. In general, these operators take a single hypergeometric
function into a sum of two hypergeometric functions which cannot be reexpressed as a single hypergeometric function except in degenerate cases where one obtains the shape invariant special cases
discussed earlier, Explicitly, one has for the unnormalized eigenfunctions of Hz, the partner to the
original Natanzon Hamiltonian:
@,
=R-4Zfi~2(1-Z)6PZ2[(1
-z)[~n-/$)-Z(Sn-S())]F(-n,a,-n;l+p,;z)
-22(1-z)n(n+1+p,+&)F(-n+1,a,-n+1;2+p,;z)1
and for the partner potential
(266)
v,=E~+[(~O+SO)(~O+~~+2)2(2-1)+(~~-1)(1-z)+(S~-1)Z+11/R
+ [a-
[c,(3Po+~o)
+co(Po+3~o)a(l
-Po-~o)l/Mz
+4Po-~o>l/[z(z
- z2)/R2.
- 111
- 1)l
(267)
By using the hierarchy of Hamiltonians one can construct in the usual manner more and more
complicated potentials which are ratios of higher and higher order polynomials in z (as well as R)
which are isospectral to the original Natanzon potential except for the usual missing states. The wave
functions will be sums of hypergeometric functions. The arguments are identical for the Ginocchio
class with the hypergeometric functions now being restricted to being Gegenbauer polynomials. The
potentials one obtains can have multiple local minima and several of these are displayed in [ 521.
In [52] it was also shown that the series of Hamiltonians and SUSY charges form the graded Lie
algebra SE( l/l) @ SU( 2). However this algebra did not lead to any new insights. We shall also see
in Section 13 that the series of Hamiltonians form a parasupersymmetry
of order p if the original
Hamiltonian has p bound states.
As we have seen in previous sections, for quantum mechanical problems, the main implication of
SUSY is that it relates the energy eigenvalues, eigenfunctions and phase shifts of two supersymmetric
partner potentials V, (x) and V,(x) . Combining the ideas of SUSY with the lowest order WKB method,
Comtet, Bandrauk and Campbell 1791 obtained the lowest order SWKB quantization condition in
319
case SUSY is unbroken and showed that it yields energy eigenvalues which are not only exact for
large quantum numbers IZ (as any WKB approximation scheme should in the semiclassical limit) but
which are also exact for the ground state (n = 0). We shall now show this in detail.
In lowest order, the WKB quantization condition for the one dimensional potential V(x) is [ 15 I]
J,/2m[E,-V(x)]dx=(n+l/2)fi7r,
n=0,1,2,...,
(268)
XI
where xi and x2 are the classical turning points defined by En = V(x,) = V(x2). For the potential
V, (x) corresponding to the superpotential W(x), the quantization condition (268) takes the form
2m[ EA*
-W(x)ldx
(269)
= (n + 1/2)fi7r.
Let us assume that the superpotential W(x) is formally 0( ti). Then, the W term is clearly 0( ti).
Therefore, expanding the left hand side in powers of ti gives
b
W(x)dx
J_+**=
2m[ EA -M(x)ldx+;/
SJ
(I
(n+
1/2)h
(270)
(I
where a and b are the turning points defined by Ei ) = w2 (a) = w2 (b) . The 0( ti) term in eq. (270)
can be integrated easily to yield
ii
. _,
W(x)
-sm
Jg
2
[
1
b
(271)
U*
W(x)
= W(b) = @.
(272)
For this case, the 0( hi) term in (271) exactly gives h7r/2, so that to leading order in ti the SWIG3
quantization condition when SUSY is unbroken is [79,80]
b
n=0,1,2
(273)
,...
SJ
a
Proceeding
be
condition
V,(x)
turns out to
SJ
2m [ EL2 -v(x)]dx=(n+l)tirr,
n=0,1,2
,...
(274)
320
(i) For n = 0, the turning points a and b in Eq. (273) are coincident and Eh = 0. Hence SWKB is
exact by construction for the ground state energy of the Hamiltonian Hr = ( -@/2m) d2/dx2 + V, (x).
(ii) On comparing Eqs. (273) and (274)) it follows that the lowest order SWKB quantization
condition preserves the SUSY level degeneracy i.e. the approximate energy eigenvalues computed
from the SWKB quantization conditions for V,(x) and V,(x) satisfy the exact degeneracy relation
,@) = E;2.
n+l
(iii) Since the lowest order SWKB approximation is not only exact as expected for large 12, but
is also exact by construction for it = 0, hence, unlike the ordinary WKB approach, the SWKB
eigenvalues are constrained to be accurate at both ends of allowed values of n at least when the
spectrum is purely discrete. One can thus reasonably expect better results than the WKB scheme.
6.2. Exactness of the SWKB condition for shape invariant potentials
How good is the SWKB quantization condition [Eq. (273) ]? To study this question, the obvious
first attempts consisted of obtaining the SWKB bound state spectra of several analytically solvable
potentials like Coulomb, harmonic oscillator, Morse, etc. [79,80]. In fact, it was soon shown that
the lowest order SWKB condition gives the exact eigenvalues for all SIPS [45] ! The proof of
this statement follows from the facts that the SWKB condition preserves the level degeneracy and a
vanishing ground state energy eigenvalue. For the hierarchy of Hamiltonians H() discussed in Section
3, the SWKB quantization condition takes the form
2m[Ei) - kR(ak)
- W(a,;x)]dx
= nfm.
(275)
k=l
i?(ak)
condition
(276)
k=l
as given by Eq. (275), must be exact for Hamiltonian H cs). One can now go back in sequential
manner from HcS) to H(S-) to Ht2) and H() and use the fact that the SWKB method preserves the
level degeneracy E,+,
(I) = Ei2). On using this relation n times , we find that for all SIPS, the lowest order
SWKB condition gives the exact energy eigenvalues [ 451. This is a very substantial improvement
over the usual WKB formula Eq. (268) which is not exact for most SIPS. Of course, one can
artificially restore exactness by ad hoc Langer-like corrections [ 1521. However, such modifications
are unmotivated and have different forms for different potentials. Besides, even with such corrections,
the higher order WKB contributions are non-zero for most of these potentials [ 152,153].
What about higher order SWKB contributions? Since the lowest order SWKB energies are exact
for shape invariant potentials, it would be nice to check that higher order corrections vanish order by
order in fi. By starting from the higher order WKB formalism, one can readily develop the higher
order SWKB formalism [ 831. It has been explicitly checked for all known SIPS that up to 0( ti6)
there are indeed no corrections. This result can be extended to all orders in fi [ 84,851. Conditions
on the superpotential which ensure exactness of the lowest order SWKB condition have been given
in Ref. [ 851.
321
It has been proved above that the lowest order SWKB approximation reproduces the exact bound
state spectrum of any SIP. This statement has indeed been explicitly checked for all SIPS known
until last year i.e. solutions of the shape invariance condition involving a translation of parameters
a2 = al + constant. However, it has recently been shown [ 1541 that the above statement is not true for
the newly discovered class [58-60,135] of SIPS discussed in Section 4.2.2, for which the parameters
a2 and al are related by scaling u2 = qul. What is the special feature of these new potentials that
interferes with the proof that Eq. (273) is exact for SIPS? To understand this, let us look again at
the derivation of the lowest order SWKB quantization condition. In the derivation, IV2 is taken as
O(ti) while tiW is 0( ti) and hence one can expand the integrand on the left hand side in powers
of ti. This assumption is justified for all the standard SIPS [ 1454,551 since W2 is indeed of 0( ti)
while hW is indeed of 0(/i,). One might object to this procedure since the resulting potential VI is
then h-dependent. However, in all cases, this h-dependence can be absorbed into some dimensionful
parameters in the problem. For example, consider
W(x)
= A tanhx,
(277)
Taking A such that A( A + h) is independent of ti gives the desired h-independent potential (the
additive constant is irrelevant and so can contain ti). Such a move may appear to be of limited
value since one cannot apply SWKB directly to a super-potential W which is now h-dependent.
However, because A is a free parameter, one can continue the SWKB results obtained for A (and
hence W) independent of fi over to this superpotential and so obtain an SWKB approximation for a
h-independent potential.
What about the new potentials? In the simplest of these cases, the only free parameter in the problem
(apart from q) is the combination R I a ,, on which W depends as W(x, Rlal) = fiF(mx/ti).
Incorporating different dependences on ti in Rlul will give different ones in W, V, and E,,, but F is
a sufficiently complicated function that there is no way of eliminating fi from w. This is a direct
consequence of the scaling reparameterisation u2 = qul not involving fi. If V(X; al) were independent
of ti, so would t@(x; a2) be and in taking the lowest order of the shape invariance condition one
would get w2( x; al) = w2 (x; a2), which corresponds to the harmonic oscillator. Furthermore, with
u2 = qal, W and tLW are now of a similar order in ti. The basic distinction between them involved in
deriving Eq. (273) is thus no longer valid and we are prevented from deriving the SWKB condition
for these new potentials.
We thus see that the SWKB quantization condition is not the correct lowest order formula in
the case of the new SIPS and hence it is not really surprising that Eq. (273) does not give the
exact eigenvalues for these potentials. In other words, it remains true that the lowest order SWKB
quantization condition is exact for SIPS (if the SUSY is unbroken), but only in those cases for which
the formula is applicable in the first place. It is thus still the case that the SIPS given in Refs. [ 54,551
are the only known ones for which the lowest order SWKB formula is exact and the higher order
corrections are all zero.
6.3. Comparison
Let us now compare the merits of the two schemes [ WKB and SWKB] . For potentials for which
the ground state wave function (and hence the superpotential W) is not known, clearly the WKB
approach is preferable, since one cannot directly make use of the SWKB quantization condition Eq.
322
(273). On the other hand, we have already seen that for shape invariant potentials, SWKB is clearly
superior. An obvious interesting question is to compare WKB and SWKB for potentials which are
not shape invariant but for whom the ground state wave function is known. One choice which readily
springs to mind is the Ginocchio potential given by [57]
V(x)
(1 - y2> -h2V(V + 1) +
i
(1 -AZ)
[2 - (7 - A2)y2 + 5( 1 - A2)y41
4
(278)
(279)
Here the parameters v and A measure the depth and shape of the potential respectively. The corresponding superpotential is [ 521
W(x) = (1 - A2)y(y2 - 1)/2 + ,uc,A2y
(280)
[A*(v+ 1/2)2+
(1-
l/2)
(281)
n=0,1,2,...
(282)
For the special case A = 1, one has the Rosen-Morse potential, which is shape invariant. The spectra of
the Ginocchio potential using both the WKB and SWKB quantization conditions have been computed
[ 901. The results are shown in Table 6.1. In general, neither semiclassical method gives the exact
energy spectrum. The only exception is the shape invariant limit A = 1, in which case the SWKB
results are exact, as expected. Also, for 12= 0, 1 the SWKB values are consistently better, but there is
no clear cut indication that SWKB results are always better. This example, as well as other potentials
studied in Refs. [ 86-901, support the conjecture that shape invariance is perhaps a necessary condition
for the lowest order SWKB approximation to yield the exact spectrum [ 901.
So far, we have concentrated our attention on the energy eigenvalues. For completeness, let us note
that several authors have also obtained the wave functions in the SWKB approximation [ 79,91-931.
As in the WKB method, the SWKB wave functions diverge at the turning points a and b. These
divergences can be regularized either by the uniform approximation [ 91,155] or by appropriate
retention of higher orders in ti [ 92,931.
6.4. SWKB quantization
condition
furbroken super-symmetry
The derivation of the lowest order SWKB quantization condition for the case of unbroken SUSY
is given in Section 6.1 [Eqs. (268) to (273)]. For the case of broken SUSY, the same derivation
applies until one examines the O(h) term in Eq. (271). Here, for broken SUSY, one has
W(a)
= W(b)
=@
(283)
323
Table 6.1
Comparison of the lowest order WKB and SWKB predictions for the bound state spectrum of the Ginocchio potential for
different values of the parameters A, Y and several values of the quantum number n. The exact answer is also given. Units
corresponding to fi = 2m = 1 are used throughout.
WKB
A = 0.5, u = 5.5
SWKB
Exact
value
WKB
A = 6.25, v = 5.5
SWKB
Exact
value
1
2
3
5
-6.18694
-3.07965
- 1.43039
-0.58466
-0.017014
-6.40641
-3.16971
- 1.46768
-0.60078
-0.018750
-6.40641
-3.12998
-1.43919
-0.58474
-0.016684
-1372.28112
-1228.15761
-1012.52312
-733.58861
-55.26983
- 1359.61147
-1212.7014
-999.6307
-727.7175
-70.5885
-1359.61147
-1213.8388
-1003.70301
-737.62022
- 109.5019
WKB
A = 0.5, v = 10.5
SWKB
Exact
value
WKB
A = 6.25, v = 10.5
SWKB
Exact
value
0
1
2
3
5
-24.86661
-17.09183
- 11.57071
-7.70199
-3.15100
-25.17048
-17.27213
- 11.67790
-7.76673
-3.17611
-25.17048
- 17.23352
- 11.63352
-7.72800
-3.15352
-4659.878
-4452.736
-4174.482
-3828.490
-2949.599
-4648.6161
-4438.321
-4158.478
-3812.566
-2978.741
-4648.6161
-4438.7989
-4159.4341
-3815.6502
-2947.7017
n
0
and the 0( ti) term in (271) exactly vanishes. So, to leading order in ti the SWKB quantization
condition for broken SUSY is [ 94,96 J
b
2m[ EP - Wz(x)]dx=(n+1/2)~7r,
n=0,1,2
,...
(284)
(I
As before, it is easy to obtain the quantization condition which includes higher orders in ti [96] and
to test how well the lowest order broken SWKB condition works for various specific examples. As for
the case of unbroken SUSY, it is found that exact spectra are obtained for shape invariant potentials
with broken SUSY [95]. For potentials which are not analytically solvable, the results using Eq.
(284) are usually better than standard WKB computations. Further discussion can be found in Ref.
[W.
7. Isospectral
Hamiltonians
In this section, we will describe how one can start from any given one-dimensional potential
V,(x) with TZbound states, and use supersymmetric quantum mechanics to construct an n-parameter
family of strictly isospectral potentials V,(hi, AZ, . . . , A,; x) i.e., potentials with eigenvalues, reflection
and transmission coefficients identical to those for VI (x) . The fact that such families exist has been
known for a long time from the inverse scattering approach [ 671, but the Gelfand-Levitan approach to
finding them is technically much more complicated than the supersymmetry approach described here.
Indeed, the advent of SUSY QM has produced a revival of interest in the determination of isospectral
potentials [ 65,68,69,156-1581
[ 66,70,72,75]. In Section 7.1 we describe how a one parameter
324
isospectral family is obtained by first deleting and then re-inserting the ground state of Vi (x) using
the Darboux procedure [ 64,701. The generalization to obtain an n-parameter family is described in
Section 7.2 [ 721. When applied to a reflectionless potential (Section 7.3)) the n-parameter families
provide surprisingly simple expressions for the pure multi-soliton solutions [75] of the Korteweg-de
Vries (KdV) and other nonlinear evolution equations [ 159,160,138,161,162].
7.1. One parameter family of isospectral potentials
In this subsection, we describe two approaches of obtaining the one-parameter family Vi(A, ; x) of
potentials isospectral to a given potential Vi (x) .
The first approach follows from asking the following question: Suppose V,(x) is the SUSY partner
potential of the original potential V, (x) , and let W(x) be the superpotential such that V,(x) = w2 + W
and V, (x) = w - W. Then, given V2(x) , is the original potential V, (x) unique i.e., for a given V,(x) ,
what are the various possible superpotentials l?(x) and corresponding potentials pi (x) = w - @?
Let us assume [68] that there exists a more general superpotential which satisfies
v,(x)
= P*(x)
+ W(x).
(285)
Clearly, @ = W is one of the solutions to Eq. (285). To find the most general solution, let us set
P(x)
= W(x) + 4(x>
(286)
= 1 + 2wy
(287)
whose solution is
+ Al].
Here
Z,(x)
Jtg(X)dX.
(289)
-cm
hi is a constant of integration and cc/,(x) is the normalized ground state wave function
w2( x) - W(x) . Thus the most general p(x) satisfying Eq. (285) is given by
P(x)
= W(x)
+ $ln[Zi(x)
AlI,
of Vi(x) =
(290)
-P(x)
=v(x)
-2$ln[&(x)
+A,]
(291)
325
i.e., it has the same eigenvalues as V,(x) except for the bound state at energy El. The next step is to
reinstate a bound state at energy El.
Although the potential V2 does not have an eigenenergy El, the function 1/I,$, satifies the Schrodinger
equation with potential V, and energy El. The other linearly independent solution is J_, @IT(x) d~/@~.
Therefore, the most general solution of the Schrijdinger equation for the potential V, at energy El is
@l (AI
> =
(11
Al I/h
(292)
Now, starting with a potential V,, we can again use the standard SUSY (Darboux)
add a state at El by using the general solution @i (Ai ),
= V, -2$lnG1(A1).
fi(A,)
procedure
to
(293)
The function I/@ (Ai ) is the normalizable ground state wave function of fi (Al ), provided that A,
does not lie in the interval -1 5 Al 5 0. Therefore, we find a one-parameter family of potentials
q (Ai ) isospectral to Vi for A, > 0 or Ai < -1.
fi(A1)=&-2-$ln($1@1(Ai))=C;-2
The corresponding
&(A,;x)
$ln(*i
+ Al).
(294)
= l/MA,).
(295)
Note that this family contains the original potential VI. This corresponds to the choices hi -+ &co.
To elucidate this discussion, it may be worthwhile to explicitly construct the one-parameter family
of strictly isospectral potentials corresponding to the one dimensional harmonic oscillator [ 701. In
this case
W(X) = ;X
(296)
so that
v,(x)
$2 - ;.
The normalized
(297)
of V,(x) is
9%(x) = (z)4exp(-WX2/4)
Using Eq. (289)
(298)
it is now straightforward
Zi (x) . We get
00
erfc(x)
= -?e-dt.
s
J;;x
(299)
Using Eqs. (294) and (295)) one obtains the one parameter family of isospectral potentials and
the corresponding wave functions. In Figs. 7.1 and 7.2, we have plotted some of the potentials and
wave functions for the case o = 2. We see that as A1 decreases from 00 to 0, ?, starts developing a
326
20
10
-10
0
-2
-4
Fig. 7.1. Selected members of the family of potentials with energy spectra identical to the one dimensional harmonic
oscillator with o = 2. The choice of units is ti = 2m = 1. The curves are labeled by the value Al, and cover the range
0 < hr 5 cc. The curve AI = cc is the one dimensional harmonic oscillator. The curve marked Al = 0 is known as the
Pnrsey potential [ 661 and has one bound state less than the oscillator.
0.25
0.00
-4
-2
Fig. 7.2. Ground state wave functions for all the potentials shown in Fig. 7.1, except the Pursey potential.
[651.
7.2. Generalization
to n-parameter
isospectral family
The second approach discussed in the previous subsection can be generalized by first deleting
all 12 bound states of the original potential VI (x) and then reinstating them one at a time. Since
VI
----______
E2
ww = (12+ WA
da
----_-_-_
El
*1
~2&)
v3
v2
327
-_-------
l/@z(U
----
@l(h) = (II
t X,)/h
%(AI, h) =
----_
&WA(~L)
I/@l(h,b)
Fig. 7.3. A schematic diagram showing how SUSY transformations are used for deleting the two lowest states of a potential
VI(x) and then re-inserting them, thus producing a two-parameter (Al, AZ) family of potentials isospectral to VI(x).
one parameter is generated every time an eigenstate is reinstated, the final result is a n-parameter
isospectral family [ 721. Recall that deleting the eigenenergy El gave the potential V2(x) . The ground
state $2 for the potential V, is located at energy E2. The procedure can be repeated upward,
producing potentials V,, Vq,. . . with ground states fiJ, @J/4,
. . . at energies E3, E4, . . , , until the top
potential V,+i (x) holds no bound state (see Fig. 7.3, which corresponds to n = 2).
In order to produce a two-parameter family of isospectral potentials, we go from & to V, to V, by
successively deleting the two lowest states of V, and then we re-add the two states at E2 and El by
SUSY transformations. The most general solutions of the Schrijdinger equation for the potential V,
are given by c&( AZ) = (12 + Az)/$z at energy Ez, and A& ( Al ) at energy El (see Fig. 7.3). The
quantities Zi are defined by
x
l&(x)
#;(x')dx'.
(300)
-co
Here the SUSY operator Ai relates solutions of the potentials 4 and x+,,
Ai = -$ - (ln&).
(301)
= V, - 2$ln(Z,
Therefore,
= -g
family 4 (AZ),
+ AZ).
one-parameter
(302)
V, and &( AZ) are related by a new SUSY
+ (ln&(A2)).
(303)
@I(AI,Az)
=&AhWdAd.
The normalizable function 1/CD,(Ai, AZ) is the ground state at El of a new potential,
in a two-parameter family of isospectral systems fi (Al, A2),
(304)
which results
328
@(Add
= V, -2~ln(d,hs,(n,)m,(r,,A~~~
=V,
-2~lnO~(K+A2)~1(A,,A2))r
(305)
=-9zV2)
is
(306)
+2(~~(A,,A2)/~l(Al,h2))2+2E1.
The above procedure is best illustrated by the pyramid structure in Fig. 7.3. It can be generalized to
an n-parameter family of isospectral potentials for an initial system with IZbound states. The formulas
for an n-parameter family are
@(hi)
Ai =
= (Z+A;)/&,
i= l,...,n;
(307)
$ - (In&);
&Ai,-,h,)
(308)
= -$
+ (ln@i(Ai, . . .,A,));
(309)
Ai+j,. . .) A,)
. . .Ai+&(Aj);
. . . , A,,) = V, - 2&
(310)
ln(cGI1@2. ..(Cln@n(An>
. ..@l(Al....,&,))
(311)
+ l)sech2x,
where n is an integer, since these potentials are of special physical interest. V, holds n bound states,
and we may form a n-parameter family of isospectral potentials. We start with the simplest case
n = 1. We have V, = -2sech2x, El = -1 and til = 2-/2sechx. The corresponding l-parameter family
is
?,(A,)
= -2sech2
x+
iln(l
+ !-)
.
>
(313)
Clearly, varying the parameter A, corresponds to translations of VI (x) . As hi approaches the limits
O+ (Pursey limit) and - 1 - (Abraham-Moses
limit), the minimum of the potential moves to -KJ
and +CG respectively.
For the case n = 2, VI = -6sechLx and there are two bound states at E, = -4 and E2 = - 1.
The SUSY partner potential is V2 = -2sech2x. The ground state wave functions of V, and V2 are
and & = ifisechx.
Also, Zi = i(3tanhx
- tanh3x + 2) and 12 = i(tanhx + 1).
@II= i&sech2x
After some algebraic work, we obtain the 2-parameter family
E Cooper et a//Physics
329
,.
[3+4cosh(2x-2&)
+cosh(4x-2&)]
(hih2)=-12,cosh(3x-62-S1)+3cosh(x+82-8S1)]2
si = -;
ln(1 + t,,
i= 1,2,
As we let Ai -+ - 1, a well with one bound state at El will move in the +x direction leaving behind a
shallow well with one bound state at E2. The movement of the shallow well is essentially controlled
by the parameter AZ. Thus, we have the freedom to move either one of the wells.
It is tedious but straightforward to obtain the result for arbitrary n and get fi (Ai, AZ, . . . , A,,, x).
It is well known that one-parameter (t) families of isospectral potentials can also be obtained as
solutions of a certain class of nonlinear evolution equations [ 159,160,138,162]. These equations have
the form (q = 0, 1,2, . . .)
-ur
= (L,)qux
(314)
LJ(x)
dyf(y)
(315)
and u is chosen to vanish at infinity. [For q = 0 we simply get -ur = u,, while for q = 1 we obtain
the well studied Korteweg-de Vries (KdV) equation]. These equations are also known to possess
pure (i.e., reflectionless)
multisoliton solutions. It is possible to show that by suitably choosing
the parameters hi as functions of t in the n-parameter SUSY isospectral family of a symmetric
reflectionless potential holding II bound states, we can obtain an explicit analytic formula for the
pure n-soliton solution of each of the above evolution equations [ 751. These expressions for the
multisoliton solutions of Eq. (314) are much simpler than any previously obtained using other
procedures. Nevertheless, rather than displaying the explicit algebraic expressions here, we shall
simply illustrate the 3-soliton solution of the KdV equation. The potentials shown in Fig. 7.4 are
all isospectral and reflectionless holding bound states at El = -25/ 16, E2 = - 1, Es = -16/25 . As t
increases, note the clear emergence of three independent solitons.
In this section, we have found n-parameter isospectral families by repeatedly using the supersymmetric Darboux procedure for removing and inserting bound states. However, as briefly mentioned in
Section 7.1, there are two other closely related, well established procedures for deleting and adding
bound states. These are the Abraham-Moses procedure [ 651 and the Pursey procedure [ 661. If these
alternative procedures are used, one gets new potential families all having the same bound state
energies but different reflection and transmission coefficients. Details can be found in reference [ 711.
E Cooper
330
et al. /Physics
-5
267-385
10
Fig. 7.4. The pure three-soliton solution of the KdV equation as a function of position (x) and time (t). This solution
results from constructing the isospectral potential family starting from a reflectionless, symmetric potential with bound states
at energies Et = -25/16, EZ = - 1, E3 = - 16/25. Further details are given in the text and Ref. [75].
which is l/2
of our previous
H = ip + iW2(X)Z - @9lW(x),
(316)
Z[j,r7,q*l
[dxl[d~l[d~*lexp[--S~+
jx+vb*
functions in Euclidean
+rl*til,
space. We
(317)
where
7
SE =
dT(&C; + ;wyX>
- @*[a, - W(X)](cI)
= {ti,$>
= (V,V)
=O,
(318)
x, =
331
dx/dr.
sx= E*+ +
SUSY transformations
S$ =
--E
(-&x
+ W(x))
(319)
where E and E* are two infinitesimal anticommuting parameters. These transformations correspond to
N = 2 supersymmetry.
The path integral over the fermions can now be explicitly performed using a cutoff lattice which
is periodic in the the coordinate x but antiperiodic in the fermionic degrees of freedom at 7 = 0 and
r = T. Namely we evaluate the fermionic path integral:
(320)
n A,,
m
where
JI,(r)
= Gexp
[kdTL&,
+ I]
(321)
boundary conditions:
yields:
T
A, =
Regulating
we obtain:
i(2m-t
T
l)g
1
- T
dTW(x).
(322)
the determinant
det [-r(X)]
= cash Idry.
0
is zero
(323)
332
Rewriting the cash as a sum of two exponentials we find, as expected that Z is the sum of the
partition functions for the two pieces of the supersymmetric Hamiltonian. Namely when the external
sources are zero:
Tr eeH1r + Tr eVHzr g Z- + Z+.
(324)
For the case when SUSY is unbroken, only the ground state of Hi contributes
have:
Z* =
as T
00
We also
(325)
where
TdT
r;:*!y).
p=
J0
A related path integral is obtained for the noise averaged correlation functions coming from a
classical stochastic equation, the Langevin equation. If we have the stochastic differential equation
k = W(x(7))
+ V(T)
(326)
where ~(7) is a random stirring forcing obeying Gaussian statistics, then the correlation functions of
x are exactly the same as the correlation functions obtained from the Euclidean quantum mechanics
related to the Hamiltonian Hi. To see this we realize that Gaussian noise is described by a probability
functional:
(327)
normalized
so that:
J
.I
1,
DW[ql=
~?7fmlf(am')
The correlation
(X(?)XCQ)
functions
hP[rllf(T)
= 03
=Fo&T - 7).
averaged over the noise are:
. . .>= 1 hP[rllX(?)X(Q)
***
(328)
where we have in mind first solving the Langevin equation explicitly for x( r] (7) ) and then averaging
over the noise as discussed in Section 2.2. Another way to calculate the correlation function is to
change variables in the functional integral from v to x.
(329)
E Cooper et al/Physics
333
= exp/
dtTrln ( [$
function
- W(x(r))]S(r-
7)
>
obey causality,
d7.
(331)
(332)
Choosing
[i!dTW(x)]
=exp
Detl$$$i
conditions
(333)
F0 = ii so that
P[T]
[--ii_%]
=Nexp
=Nexp
T
=
Nexp
-A
dT(k2 + W2(x))
J
0
[--$/dT(i-
,1
W(x))2]
(334)
we find that the generating functional for the correlation functions is exactly the generating functional
for the correlation functions for Euclidean quantum mechanics corresponding to the Hamiltonian Hr :
= N/D[x]
Z[j]
exp 1-f
L
jdT(i2
+ W2(x)
- W(x) - 2j(r)x(7))
.1
(335)
Thus we see that we can determine the correlation functions of x for the Hamiltonian HI by either
evaluating the path integral or solving the Langevin equation and averaging over Gaussian noise.
An equation related to the Langevin equation is the Fokker-Planck equation, which defines the
classical probability function PC for the equal time correlation functions of Hr. Defining the noise
average:
PC(Z) = @(z
one obviously
- x(t&
= JDrlS(z
- x(t))P
has:
dzzP,(z,
t) =
JD?7[x(t)lP[771 = (x).
[VI
(336 >
334
$0 f$
dt
For an equilibrium
equation:
+ &[W(z)P(e,t))l.
distribution
(337)
Ej(z)dz = 1.
=Nexp
equation, we obtain
(-2ily(v)dy)
=400(z12.
(338)
Thus at long times only the ground state wave function contributes (we are in Euclidean space)
and the probability function is just the usual ground state wave function squared. We see from this
that when SUSY is broken, one cannot define an equilibrium distribution for the classical stochastic
system.
A third path integral for SUSY QM is related to the Witten index. As we discussed before, one
can introduce a fermion number operator via
1 -
CT3
nF=-----=
I-
M~91
(339)
Thus
(-ly=
[$,t)+]
=u3.
(340)
The Witten index is given by D = Tr( -l)F. As we discussed earlier, the Witten index needs to be
regulated and the regulated index is defined as:
A(,@ =Tr(-l)Fe-PH=Tr(e-PH
-eePN2).
(341)
In Section 2.2 we showed how to determine A(p) using heat kernel methods and how it was useful
in discussing non-perturbative breaking of SUSY. Here we will show that the Witten index can also
be obtained using the path integral representation of the generating functional of SUSY QM where
the fermion determinant is now evaluated using periodic boundary conditions to incorporate the factor
(- l)F. It is easy to verify a posteriori that this is the case. Consider the path integral:
P
A(P) =
where
(342)
LE =
ix;+ ;w2 -
335
P*[a, - W(x)]%?
To incorporate the (- 1) F in the trace, one changes the boundary conditions for evaluating the fermion
determinant at r = 0, p to periodic ones:
The path integral over the fermions can again be explicitly performed using a cutoff lattice which is
periodic in the fermionic degrees of freedom at 7 = 0 and r = p. We now impose these boundary
conditions on the determinant of the operator [a, - W(x)] using eigenvectors which are periodic.
We again have
det[& - W(x)]
n A,,.
m
conditions:
(343)
Regulating the determinant
we obtain:
is zero
(344)
Again rewriting the sinh as a sum of two exponentials
regulated Witten index:
d(p)
= Z- - 2, =Tre-PH
we find, as expected
- TreFPH2.
of supersymmetric
(345)
quantum mechanics
One can think of SUSY QM as a degenerate case of supersymmetric field theory in d = 1 in the
superspace formalism of Salam and Strathdee [ 1671 (This idea is originally found in unpublished
lecture notes of S. Raby [ 1661). superfields are defined on the space (x,; 9,) where x is the space
coordinate and 8, are anticommuting spinors. In the degenerate case d = 1 the field is replaced by
x(t) so that the only coordinate is time. The anticommuting variables are 19and 0 where
(0,
e*}= (6,6}
= [B, t] = 0.
SUSY transformation:
O=O+e,
(346)
~*=O*+E*.
can be parametrized
by
336
,i(c*Q*+Qd
then from
SA = i[e*Q* + QE, A]
(347)
Q* = -ids* - Od,.
(348)
[Q, H] = 0.
(349)
The Lagrangian in superspace is determined as follows. A superfield made up of x and 19and 8* can
at most be a bilinear in the Grassman variables:
4,(x, 19,0*>= X( t> + i&b(t) - i$*O* + ee*D(t).
(350)
ie*ar
or in component form:
D& = irC,- O*D - iO*k + ti*Ot,b,
(351)
and
De* =
as. - iea,
or in component form:
[D&l
(352)
s=
Jdtde*de(;p8+(2
- f(4)).
(353)
Again the expansion in terms of the Grassman variables causes a Taylor expansion of f to truncate
at the second derivative level. Integrating over the Grassman degrees of freedom using the usual path
integral rules for Grassman variables:
/ede=Jeve*=l,
/de=/de*=o,
one obtains
S=
dt(~12++*[a,-f(x)]~+;D2+Df(x)).
S=
J
- W(x)]~ - $w>
(354)
= W(x)
(355)
A more complete
9. Perturbative
discussion
337
The framework of supersymmetric quantum mechanics has been very useful in generating several
new perturbative methods for calculating the energy spectra and wave functions for one dimensional
potentials. Four such methods are described in this section.
In Sections 9.1 and 9.2, we discuss two approximation methods (the variational method and the Sexpansion) for determining the wave functions and energy eigenvalues of the anharmonic oscillator
making use of SUSY QM. Section 9.3 contains a description of a SUSY QM calculation of the
energy splitting and rate of tunneling in a double well potential. The result is a rapidly converging
series which is substantially better than the usual WKB tunneling formula. Finally, in Section 9.4, we
describe how the large N expansion (N = number of spatial dimensions) used in quantum mechanics
can be further improved by incorporating SUSY.
9.1. Variational approach
The anharmonic oscillator potential V(x) = gx4 is not exactly solvable. To determine the superpotential one has to first subtract the ground state energy EO and solve the Riccati equation for
W(x):
v, (x) = gx4 - EOG W2 - W,
(356)
Once the ground state energy and the superpotential is known to some order of accuracy, one can
then determine the partner potential and its ground state wave function approximately. Then, using
the SUSY operator
-$ - W(x)
one can construct the first excited state of the anharmonic oscillator in the usual manner. Using the
hierarchy of Hamiltonians discussed in Section 3, one can construct from the approximate ground
state wave functions of the hierarchy and the approximate superpotentials W, all the excited states of
the anharmonic oscillator approximately.
First let us see how this works using a simple variational approach. For the original potential, we
can determine the optimal Gaussian wave function quite easily. Assuming a trial wave function of
the form
(357)
we obtain
(H)=
(;+gx4)=;+s.
(358)
338
(In this subsection, we are taking m = 1 in order to make contact with published numerical results).
Minimizing the expectation value of the Hamiltonian with respect to the parameter p yields
E0 = ( 94/3g/3
( 2)/3g/3
4
This is rather good for this crude approximation since the exact ground state energy of the anharmonic
oscillator determined numerically is EO = 0.668g/3 whereas (4)
3 4/3 = 0.681. The approximate potential
W resulting from this variation calculation is
d log @o
= 2px,
W(x) = - dx
(359)
(360)
(361)
Since V2c differs from Vi, by a constant, the approximate ground state wave function for V2is given
by Eq. (357). The approximate ground state energy of the second potential is now
(H2) = (Ol$ + V2G(O)?
= H1 + 4p.
(362)
Thus we have approximately that the energy difference between the ground state and first excited
state of the anharmonic oscillator is
E, - Eo=4p=4($)3g3.
This method can be used to find all the excited state wave functions and energy levels of the
anharmonic oscillator by using the methods discussed in Ref. [ 1681.
Let us now look at a more general class of trial wave functions. If we choose for the trial ground
state wave functions of the hierarchy of Hamiltonians,
(364)
we obtain much better agreement for the low lying eigenvalues and eigenfunctions. It is convenient
in this case to first scale the Hamiltonian for the anharmonic oscillator,
H=
-;-$
+gx4,
(365)
339
by letting x + x/gi6 and H -+ g 1/3H. Then we find the ground state energy of the anharmonic
oscillator and the variational parameters p1 and YZ~
by forming the functional
Eo(PIvm)
(eel
r$
(366)
++o).
LEO
- 0.
anl-
(367)
oscillator is
(368)
Minimizing
this expression,
E. = 0.66933,
~1~= 1.18346,
p1 = 0.666721.
(369)
This ground state energy is to be compared with a numerical evaluation which yields 0.667986. Since
the trial wave function for all ground states is given by Eq. (364), the variational superpotential for
all k is
wk, = n&Cj2nk-1(&) +.
(370)
oscillator, we consider
&k+, = ;&A;,
(371)
%(Pkv
- $
+ W,:, + W;kI~;k+)).
energy splittings
(372)
nk)
z(Zn,-1
4
2pk
1)
(373)
(fi)
in p analytically
to perform numerically.
340
Table 9. I
Comparison
Level
En - En-1
variation
exact
0
1
2
1.183458
0.995834
1.000596
0.666721
0.429829
0.435604
0.669330
1.727582
2.316410
0.667986
1.725658
2.303151
The results for the variational parameters and for the energy differences are presented in Table 9.1
for the first three energy eigenvalues and compared with a numerical calculation, based on a shooting
method.
9.2. S expansion method
In this method [78] we consider the anharmonic oscillator as an analytic continuation from the
harmonic oscillator in the paramater controlling the anharmonicity. That is we consider simultaneously
potentials of the form
v, (x) = M2+Jx2+2S-C(S)
= W2(x,S) - w,
(374)
where M is a scale parameter, S measures the anharmonicity, and C is the ground state energy of the
anharmonic oscilator. C is subtracted as usual from the potential so it can be factorized. The standard
anharmonic oscillator corresponds to 6 = 1 and A4= (2g) I3 . To approximately determine W(x) from
V,(x) we assume that both W(x) and V,(x) have a Taylor series expansion in S. Thus we write:
O P[ln(Mx2)]
n,
v,(x) = M2x2x
n=o
O
- c2E,S,
IF0
(375)
where E, corresponds to the order Taylor expansion of the dependence of the ground state energy on
the parameter 6. We assume
W(x) =
2n=o6W(,,(xl
(376)
and insert these expressions in Eq. (374) and match terms order by order. At lowest order in S the
problem reduces to the supersymmetric harmonic oscillator. We have:
W2
= M2x2 - 2E O?
0 - W
0
(377)
whose solution is
W,(x)
=Mx,
EO= ;M
(378)
dx
(379)
341
which is to be solved with the boundary conditionW,(O) = 0. The order 6 contribution to the energy
eigenvalue El is determined by requiring that the ground state wave function be square integrable.
Solving for WI we obtain
x
W (x)
-eMx2 dye-y*
(380)
(381)
Writing M = (2g) j3, we find that the first two terms in the S expansion for the ground state energy
are
E = ;(2g)3[
1 + ;$(3/2)6].
(382)
At S= 1, we get
E = 0.6415g3.
(383)
A more accurate determination of the ground state energy can be obtained by calculating up to order
S* and then analytically continuing in S using PadC approximants. This is discussed in Ref. [ 1691.
9.3. Supersymmetry
Supersymmetric quantum mechanics has been profitably used to obtain a novel perturbation expansion for the probabililty of tunneling in a double well potential [ 341. Since double wells are widely
used in many areas of physics and chemistry, this expansion has found many applications ranging
from condensed matter physics to the computation of chemical reaction rates [ 3 l-33,170-180].
In
what follows, we shall restrict our attention to symmetric double wells, although an extension to
asymmetric double wells is relatively straightforward [ 1811.
Usually, in most applications the quantity of interest is the energy difference t - El - E. between
the lowest two eigenstates, and corresponds to the tunneling rate through the double-well barrier. The
quantity t is often small and difficult to calculate numerically, especially when the potential barrier
between the two wells is large. Here, we show how SUSY facilitates the evaluation of t. Indeed, using
the supersymmetric
partner potential V2(x), we obtain a systematic, highly convergent perturbation
expansion for the energy difference t. The leading term is more accurate than the standard WKB
tunneling formula, and the magnitude of the nonleading terms gives a reliable handle on the accuracy
of the result.
First, we briefly review the standard approach for determining t in the case of a symmetric,
one-dimensional
double-well potential, V, (x), whose minima are located x = &x0. We define the
E Cooper
342
et al./Physics
267-385
partner potential
h(x).
depth, D, of V, (x) by D G V, (0) - V, (x0). An example of such a potential is shown in Fig. 9.1.
For sufficiently deep wells, the double-well structure produces closely spaced pairs of energy levels
lying below V, (0). The number of such pairs, n, can be crudely estimated from the standard WKB
bound-state formula applied to V, (x) for x > 0:
W7T=
J
0
-[v,(o)
- v,(x)p2dx,
(384)
where x, is the classical turning point corresponding to energy VI (0) and we have chosen units where
h = 2m = 1. We shall call a double-well potential shallow if it can hold at most one pair of bound
states, i.e., yt 2 1. In contrast, a deep potential refers to n >_ 2.
The energy splitting t of the lowest-lying pair of states can be obtained by a standard argument
[ 1261. Let x(x) be the normalized eigenfnnction for a particle moving in a single well whose
structure is the same as the right-hand well of V,(x) (i.e., x > 0). If the probability of barrier
penetration is small, the lowest two eigenfunctions
of the double-well potential Vj(x) are well
approximated by
G;;;(x)
By integration
= [X(X) f
xc-XNIJZ.
of Schrbdingers
(385)
t E El - I30 = 4/Y(O)x(O),
where the prime denotes differentiation with respect to x. This result is accurate for deep potentials,
but becomes progressively worse as the depth decreases. Use of WKB wave functions in Eq. (386)
yields the standard result:
tWm = {[2Vi(xo)]/2/~}exp
-271&(x)
(
- Vi(xo)]/*dx
(387)
The same result can also be obtained via instanton techniques [ 1821.
Using the supersymmetric
formulation of quantum mechanics for a given Hamiltonian, Hi =
-d2/dx2 + VI (x), and its zero-energy ground state wave function $o(x), we know that the supersymmetric partner potential V2(x) is given by
w/w
= w*(x)
343
(388)
= --$~/t,ho we can write
5 dW/dx.
(389)
From the discussion of unbroken SUSY in previous sections, we know that the energy spectra of
the potentials V2 and VI are identical, except for the ground state of VI which is missing from the
spectrum of VT [ lo]. Hence, for the double-well problem, we see that if VI (x) is shallow (i.e.,
only the lowest two states are paired), then the spectrum of V2 is well separated. In this case, V2
is relatively structureless and simpler than V,. Previous papers [ 3 I-331 have implicitly treated just
the case of shallow potentials, and, not surprisingly, have found that the use of SUSY simplifies the
evaluation of the energy difference t. In contrast, let us now consider the case of a deep double well
as shown in Fig. 9.1. Here, the spectrum of V2 has a single unpaired ground state followed by paired
excited states. In order to produce this spectrum, V2has a double-well structure together with a sharp
a-- function like dip at x = 0. This central dip produces the unpaired ground state, and becomes
sharper as the potential VI (x) becomes deeper.
As a concrete example, we consider the class of potentials whose ground state wave function is
the sum of two Gaussians, centered around &x0,
#o(x)
N e
-(x-x0)Z + e-(x+x0)**
(390)
The variables x and x0 have been chosen to be dimensionless. The corresponding superpotential
W(x), and the two supersymmetric partner potentials V,(x) and V,(x), are given respectively by
W(x)
= 2[x - x0 tanh(2xxo)],
V,,,(x) =4[x-xotanh(2xxo)]*f2[1
(391)
-2xisech2(2xxo)].
The minima of V,(x) are located near &x0 and the well depth (in the limit of large x0) is
We illustrate the potentials V, (x) and V,(x) in Fig. 9.2 for the two choices x0 = 1.0 and
We see that in the limit of large x0, for both VI (x) and V*(x), the wells become widely
and deep and that V,(x) develops a strong central dip.
The asymptotic behavior of the energy splitting, t, in the limit x0 + 00 can be calculated
(386), with x(x) given by one of the (normalized) Gaussians in Eq. (390). We find that
t -+ 8x0 (2/q)
12e-2t.
(392)
D N 4x;.
x0 = 2.5.
separated
from Eq.
(393)
The same result can be obtained by observing that V,(x) --f 4() x 1 -x0)* as x0 --f 00. This potential
has a well known [ 1831 analytic solution, which involves solving the parabolic cylindrical differential
equation. After carefully handling the boundary conditions, one obtains the separation of the lowest
two energy levels to be 8x0 (2/r) /* exp( -2x;),
in agreement with Eq. (393).
We now turn to the evaluation of t via the ground state energy of the supersymmetric
partner
potential V2(x). In general, since V,(x) is not analytically solvable, we must solve an approximate
problem and calculate the corrections perturbatively. The use of SUSY, coupled with the observation
that the magnitude of t is in general small, allows us to construct a suitable unperturbed problem.
Consider the Schrijdinger equation for V*(x) and E = 0. From supersymmetry
[Eq. (388) I we see
immediately that l/Go is a solution. Since t is small, we expect this solution to be an excellent
344
-10 _
---_-
+=1.0
xp2.6
-20 -
Fig. 9.2. Supersymmetric partner potentials VI(X) and Vz(x) corresponding to two choices of the parameter x0 for the
potentials given in Eq. (392). For a detailed discussion, see the text and Ref. [34].
approximation to the correct eigenfunction for small values of x. However, l/& is not normalizable
and hence is not acceptable as a starting point for perturbation theory. One possibility is to regularize
the behavior artificially at large 1x1 [ 3 11. This procedure is cumbersome and results in perturbation
corrections to the leading term which are substantial. Instead, we choose for our unperturbed problem
the second linearly independent solution of the Schrodinger equation given by [ 1841
(394)
and $(x) = 4(-x)
for x < 0. Clearly, +( x ) is well behaved at x = foe and closely approximates
l/(cro at small x; thus we expect it to be an excellent approximation of the exact ground state wave
function of V,(x) for all values of x. The derivative of 4(x) is continuous except at the origin, where,
unlike the exact solution, it has a discontinuity 4],,+ - 410_ = -2$,,(O). Hence 4(x) is actually a
zero-energy solution of the Schrodinger equation for a potential Vo( x) given by
v,(x)
= v,(x)
- 4~;(O)~(x)~
(395)
where we have assumed that +0(x) is normalized. We calculate the perturbative corrections to
the ground state energy using AV = +4# (0) 6( x ) as the perturbation. Note that the coefficient
multiplying the Munction is quite small so that we expect our perturbation series to converge rapidly.
For the case of a symmetric potential such as V,(x), the perturbative corrections to the energy
arising from AV can be most simply calcualted by use of the logarithmic perturbation-theory
[ 1851
formulation of the usual Rayleigh-Schrodinger
series. The first and second order corrections to the
unperturbed energy E = 0 are
E(l) = -
25(O)
Ec2 =
x
-2[[E;;;)]2~x(
(396)
w
I
PO
_.-.-.-.-.
345
ImlOr&r
- - - - - - 2ti order
I
4.3
Fig. 9.3. The energy splitting t = El - & as a function of the separation 2x0 of the superposed Gaussians in the ground
state wave function @O(X). This figure, taken from Ref. [34], shows the remarkable accuracy of the SUSY-based energy
splitting computations for a double well potential.
For our example, we numerically evaluate these corrections in order to obtain an estimate of t.
The results are shown in Fig. 9.3 for values of x0 < 2. Estimates of t correct to first, second, and
third order calculated from logarithmic perturbation theory are compared with the exact result for
V,, obtained by the Runge-Kutta method. The asymptotic behavior of t given by Eq. (393) is also
shown. This asymptotic form can also be recovered from Eq. (396) by a suitable approximation of
the integrand in the large-x0 limit. Even for values of x0 5 l/d,
in which case Vi(x) does not
exhibit a double-well structure, the approximation technique is surprisingly good. The third-order
perturbative result and the exact result are indistinguishable for all values of x0.
In conclusion, we have demonstrated how SUSY can be used to calculate t, the energy splitting for
a double-well potential. Rather than calculating this splitting as a difference between the lowest-lying
two states of V, (x), one can instead develop a perturbation series for the ground state energy t of the
partner potential V,(x) . By choosing as an unperturbed problem the potential whose solution is the
normalizable zero-energy solution of V2(x) , we obtain a very simple &-function perturbation which
produces a rapidly convergent series for t [ 341. The procedure is quite general and is applicable to
any arbitrary double-well potential, including asymmetric ones [ 1811. The numerical results are very
accurate for both deep and shallow potentials.
9.4. Supersymmetry
The large-N method, where N is the number of spatial dimensions, is a powerful technique for
analytically determining the eigenstates of the Schrodinger equation, even for potentials which have
no small coupling constant and hence not amenable to treatment by standard perturbation theory
346
[ 186-188,190-195,200].
A slightly modified, physically motivated approach, called the shifted
large-N method [ 196-1991 incorporates exactly known analytic results into l/N expansions, greatly
enhancing their accuracy, simplicity and range of applicability. In this subsection, we will descibe
how the rate of convergence of shifted l/N expansions can be still further improved by using the
ideas of SUSY QM [ 771.
The basic idea in obtaining a l/N expansion in quantum mechanics consists of solving the
Schrodinger equation in N spatial dimensions, assuming N to be large, and taking l/N as an
artificially created expansion parameter for doing standard perturbation theory. At the end of the
calculation, one sets N = 3 to get results for problems of physical interest in three dimensions.
For an arbitrary spherically symmetric potential V(r) in N dimensions, the radial Schrodinger
equation contains the effective potential
Kff(r> = V(r) +
(k-W-3)fL2,
k=N+21
(397)
8mr2
It is important to note that N and I always appear together in the combination k = N + 21. This
means that the eigenstates, which could in principle have depended on the three quantities N, 1, IZ, in
fact only depend on k and n, where n is the radial quantum number which can take values 0, 1,2, . . .
One now makes a systematic expansion of eigenstates in the parameter l/&, where i; = k - a. Of
course, for very large values of N, the two choices z and k are equivalent. However, for N = 3
dimensions, a properly chosen shift a produces great improvements in accuracy and simplicity. For
power law potentials V(r) = Ar, the shift parameter a is taken to be [ 1961
a=2-(2n+l)JYT-2.
This choice is motivated by requiring the l/R expansion to yield the exact eigenvalues for the cases
v = -1 (Coulomb potential) and v = -2 (harmonic oscillator), which are not only of physical
interest but also analytically solvable in N-dimensions. For general sperical potentials V(r), the shift
a is chosen so as to make the next to leading contribution in the l/i expansion vanish [ 1971. At
small values of r, the yt = 0 wave function fiO(r) has the behavior r(k-)/2. If one sets
@0(r)
= r
w)/2@o(r)
(3%
where @o(r) is finite at the origin, then Eq. (398) readily gives the supersymmetric
of V&(r) to be
Vi(r)
= V(r)
(kf
l)(k8mr2
l)ti2
ti2 d2
- --2
ln@0(r).
partner potential
(399)
V2(r) and V,,(r) have the same energy values [except for the ground state]. However, large-N
expansions with the partner potential V2(r) are considerably better since the angular momentum
where k = k + 2. So, effectively, one
barrier in Eq. (399) is given by (k - 1) (k - 3)h2/8mr2,
is working in two extra spatial dimensions! Thus, for example, in order to calculate the energy of
the state with quantum numbers k, II of V&(r) one can equally well use k = k + 2, II - 1 with
V2(r) . To demonstrate this procedure, let us give an explicit example. Using the usual choice of units
ti = 2m = 1, the s-wave Hulthen effective potential in three dimensions and its ground state wave
function are:
t$(r) =
-p--:ir
+C2;Q2,
&)(r)
(1 -
e-~r)e-(2-w2,
347
(400)
where the parameter 6 is restricted to be less than 2. The supersymmetric partner potential is
2S2ep8
(401)
As r tends to zero, &H goes like 2r-2, which as mentioned above, corresponds to the angular
momentum barrier (k - 1) (k - 3) h2/8mr2 for k = 5 ( N = 5, E= 0). Let us compute the energy of
the first excited state of Kg<r). For the choice S = 0.05, the exact answer is known to be 0.748125
[ 2011. The results upto leading, second and third order using a shifted l/N expansion for x$
are 0.747713, 0.748127 and 0.748125 . The corresponding values using the supersymmetric partner
potential 4 are all 0.748125! It is clear that although excellent results are obtained with the use of the
shifted l/N expansion for the original potential V$(r) in three dimensions, even faster convergence
is obtained by using the supersymmetric partner potential, since we are now effectively working
in five dimensions instead of three. Thus, SUSY has played an important role in making a very
good expansion even better [ 771. In fact, for many applications, considerable analytic simplification
occurs since it is sufficient to just use the leading term in the shifted 1/N expansion for V2(r) . Other
examples can be found in Ref. [ 771.
10. Pauli equation and supersymmetry
So far, we have discussed the concept of SUSY for Schrijdinger Hamiltonians in one-dimension
and for central potentials in higher dimensions which are essentially again one-dimensional problems.
In this section we shall show that the concept of SUSY can also be applied to some geniune
two-dimensional problems. In particular we show that the Pauli Hamiltonian which deals with the
problem of a charged particle in external magnetic field can always be put in SUSY form provided the
gyromagnetic ratio is equal to two. It is worth emphasizing here that not only the uniform magnetic
field problem (i.e. the famous Landau level problem) but the nonuniform magnetic field problems
can also be put in SUSY form. Using the concepts of SUSY and shape invariance we show that some
of the nonuniform magnetic field problems can be solved analytically.
The Pauli Hamiltonian for the motion of a charged particle in external magnetic field in two
dimensions is given by ( ti = 2m = e = 1)
H = (px +
(402)
It is easily seen that this H along with the supercharges Q and Q2 defined by [35,39]
Q' =
-&P, +A&,
Q2 =
+ (p, + A,bJ,
+
Ay)gyl
(403)
{Q",QP}=H6"P,[H,Q-]
=O;
&p=l,2.
(404)
348
E Cooper
et al./Physics
267-385
It is interesting to note that SUSY fixes the value of g. The above Hamiltonian has an additional
0( 2) @O( 2) symmetry coming from u, and an 0( 2) rotation in the A, A2 plane (A E px + A,; A2 s
py + A!,). Let us first consider the problem in an asymmetric gauge i.e.
A+,
y) = 0, A,(x, y) = W(y)
where W(y)
(405)
H = (P, + W(Y))*
+ P; - W(yh.
(406)
6 can be factorized
$(X, Y) = @V(Y)
(407)
-$ +
(W(Y)
as
+k>*- W(y>a
$(y>
= E$(~)
where u( = f 1) is the eigenvalue of the operator gZ. Thus we have reduced the problem to that
of SUSY in one dimension with superpotential W(y) + k where W(y) must be independent of k.
This constraint on W(y) strongly restricts the allowed forms of shape invariant W(y) for which
the spectrum can be written down algebraically. In particular from Table 4.1 we find that the only
allowed forms are (i) W(y) = w,y + cl (ii) W(y) = atanhy + cl (iii) W(y) = atany + cl (iv)
W(y) = cl - c2 exp (-y) for which W(y) can be written in terms of simple functions and for which
the spectrum can be written down algebraically [202,101]. In particular when
W(Y)
(409)
= @cY + Cl
which corresponds
are given by
to the uniform magnetic field, then the energy eigenvalues known as Landau levels
E,=(2n+l+a)w,;
n=0,1,2,...
(410)
Note that the ground state and all excited states are infinite-fold degenerate since E,, does not depend
on k which assumes a continuous sequence of value (--00 5 k 5 co).
The magnetic field corresponding to the other choices of W are (ii) B = -asech*y
(iii)B =
--a set* y ( 7 < y 5 T ) (iv) B = +c2 exp ( -y ) and as mentioned above, all these problems can be
solved algebrically by using the results of Section 4.
Let us now consider the same problem in the symmetric gauge. We choose
A, = ocyf(p)
A,, = -w,xf(p)
(411)
- wG(d.
(412)
F: Cooper et d/Physics
349
ccl(~, 4) = R(p)eVfi
(414)
+ w,Pf(P))u+
?n2 - l/4
p2
equation
for R(p)
R(P)
(415)
= -WP)
where a( = &l) is the eigenvalue of the operator crZ. On comparing with Table 4.1, it is easily
checked that there is only one shape invariant potential (f(p)
= 1) for which the spectrum can
be written down algebraically. This case again corresponds to the famous Landau level problem i.e.
it corresponds to the motion of a charged particle in the x - y plane and subjected to a uniform
magnetic field (in the symmetric gauge) in the z-direction. The energy eigenvalues are
E,,=2(n+m+~m~)w,;
n=0,1,2,...
(416)
so that all the states are again infinite-fold degenerate. It is worth noting that the nonuniform magnetic
fields can also give this equi-spaced spectrum [203]. However, they do so only for one particular
value of m while for other values of m, the spectrum is in general not equi-spaced.
The fact that in this example there are infinite number of degenerate ground states with zero energy
can be understood from the Aharonov-Casher theorem [ 204 J which states that if the total flux defined
by @ = J B, dxdy = n + ~(0 < E < 1) then there are precisely n - 1 zero energy states. Note that in
our case @ is infinite.
11. Supersymmetry
There have been many applications of SUSY QM in the context of the Dirac equation. In view of
the limitations of space, we shall concentrate on only a few of these applications [ 1001. In particular,
we discuss the supersymmetric structure of the Dirac Hamiltonian (HD) and show how the methods
used to obtain analytical solutions of the Schrodinger equation can be extended to the Dirac case.
First of all, we consider the Dirac equation in 1 + 1 dimensions with Lorentz scalar potential $J( x) .
We show that whenever the one-dimensional
Schrodinger equation is analytically solvable for a
potential V(x), then there always exists a corresponding Dirac scalar potential problem which is also
analytically solvable [ 1011. It turns out that, on the one hand, $(x) is essentially the superpotential
of the Schrodinger problem and on the other hand, it can be looked upon as the kink solution of a
scalar field theory in l+l dimensions. Next we discuss the celebrated problem of the Dirac particle
in a Coulomb field [ 1031 and show that its eigenvalues and eigenfunctions can be simply obtained by
using the concept of SUSY and shape invariance as developed in Sections 2 and 4. We also discuss
the problem of the Dirac equation in an external magnetic field in two dimensions and show that
there is always a supersymmetry
in the problem in the massless case. We also classify a number
350
of magnetic field problems whose solutions can be algebraically obtained by using the concepts of
SUSY and shape invariance [ 1011. In addition, we show that the Euclidean Dirac operator in four
dimensions, in the background of gauge fields, can always be cast in the language of SUSY QM.
Finally, we discuss the path integral formulation of the fermion propagator in an external field and
show how the previously known results for the constant external field can be very easily obtained by
using the ideas of SUSY [ 1011.
11.I. Dirac equation with Lorentz scalar potential
The Dirac Lagrangian in 1 + 1 dimensions with a Lorentz scalar potential 4(x) is given by
.C = iJya,$
- (I;@$.
(417)
The scalar potential 4(x) can be looked upon as the static, finite energy, kink solution corresponding
to the scalar field Lagrangian
(418)
Such models have proved quite useful in the context of the phenomenon of fermion number
fractionalization [205-2071 which has been seen in certain polymers like polyacetylene. Further,
a variant of this model is also relevant in the context of supersymmetric field theories in 1 + 1
dimensions [208,209]. Note that the coupling constant in Eqs. (417) and (418) has been absorbed
in C$and V( 4) respectively.
The Dirac equation following from Eq. (417) is
Wa,$(x,
t) - @(x)#(x, t) = 0.
(419)
Let
+(x, t) = exp(-id)@(x)
(420)
( 1,
y
:,
am
= 0.
y = i(r3 =
(421)
(5 _Oi),
(422)
A+&(x) = @t(x),
(423)
where
A=
A+ = -$
+ 4(x).
(424)
351
AAte
= w2e2.
(425)
On comparing with the formalism of Section 2, we see that there is a supersymmetry in the problem and 4(x) is just the superpotential of the Schrodinger formalism. Further #,1 and ti2 are the
eigenfunctions of the Hamiltonians H- E AfA and H, = AAt respectively with the corresponding
potentials being V, (x) = +2(x) 7 4 (x) . The spectrum of the two Hamiltonians is thus degenerate
except that H_ (H+) has an extra state at zero energy so long as 4(x t *too) have opposite signs
and 4(x --f +oc) > 0( < 0). Using the results of the Sections 2 and 4 we then conclude that for
every SIP given in Table 4.1 there exists an analytically solvable Dirac problem with the corresponding scalar potential 4(x) being the superpotential of the Schriidinger problem. In particular, using
the reflectionless superpotential given by
W(X) = ntanhx
(426)
one can immediately construct perfectly transparent Dirac potentials with IZ bound states [ 2101.
Further, using the results for the SIP with scaling ansatz (a2 = qal) [58,59], one can also construct
perfectly transparent Dirac potentials with an infinite number of bound states.
11.2. Super-symmetry and the Dirac particle in a Coulomb $eld
The Dirac equation for a charged particle in an electromagnetic field is given by (e = ti = c = 1)
[ir(a, + iA,)
- rn]ti = 0.
(427)
For a central field i.e. A = 0 and Ao(x, t) = V(r), this equation can be written as [ 21 l]
(428)
where
(429)
with C# being the Pauli matrices. For central fields, this Dirac equation can be separated in spherical
coordinates and finally for such purposes as the computation of the energy levels one only needs to
concentrate on the radial equations which are given by [ 2111
G(r) + y
- (cy, - V)F = 0,
F(r)
kF
- - (a2 + V)G = 0
r
(430)
where
a1 =m+E,
cz2=m-E
(431)
and Gk is the large component in the non-relativistic limit. Of course the radial functions Gk and
Fk must be multiplied by the appropriate two component angular eigenfunctions to make up the full
four-component solutions of the Dirac equation [211]. These coupled equations are in general not
352
y=ze2
(432)
We now show that the Coulomb problem can be solved algebraically by using the ideas of SUSY
and shape invariance. To that end, we first note that in the case of the Coulomb potential, the coupled
equations (430) can be written in a matrix form as
(433)
where k is an eigenvalue of the operator -(G . L + 1) with the allowed values k = f 1, f2, f3, . . .,
and satisfies 1kl = J + i. Following Sukumar [ 1031, we now notice that the matrix multiplying l/y
can be diagonalized by multiplying it by a matrix D from the left and D-l from the right where
,-rs),
Jiq.
s=
(434)
On multiplying Eq. (433) from the left by the matrix D and introducing
leads to the pair of equations
(;+;>
P
(435)
where
(:)=D(z)
(436)
and
A=d_:+r
dp
At=_?-
dp
-;+;.
(437)
Thus we can easily decouple the equations for p and rI? thereby obtaining
(438)
in the problem and H+ are shape invariant supersymmetric
H+(p;s,y)
On comparing
u2=s+l,
= H-(p;s+
l,r> +;
(s:
1>2
(439)
R(a2) = r _ r
a:
a:
(440)
(i,z- ($)
E;-
of H- are given by
112
[l+
(441)
k=2
Y
(,I n2)
353
n=0,1,2,...
(442)
of H- is also an eigenvalue
(443)
Using the formalism for the SIP as developed in Section 4, one can also algebraically obtain all the
eigenfunctions of P and e.
Notice that the spectrum as given by Eq. (442) only depends on ]k] leading to a doublet of states
corresponding to k = )kl and k = -(kJ for all positive n. However, for n = 0, only the negative value
of k is allowed and hence this is a singlet state.
11.3. SUSY and the Dirac particle in a magnetic$eld
Let us again consider the Dirac equation in an electromagnetic field as given by Eq. (427) but
now consider the other case when the vector potential is nonzero but the scalar potential is zero i.e.
A0 = 0, A $0. Now as shown by Feynman and Gell-Mann [ 2121 and Brown [ 2131, the solution
of the four component Dirac equation in the presence of an external electromagnetic
field can be
generated from the solution of a two component relativistically invariant equation. In particular, if @
obeys the two component equation
[(P+A)2+m2+(T.(B+iE)l~=((B+Ao)21CI
(444)
then the four component spinors that are solutions of the massive Dirac equation are generated
the two component $ via
(o.(P+A)+B-Ao+m)$
lLD=((~,~~+A)+E-Ao-m),
>
from
(445)
Thus, in order to solve the Dirac equation, it is sufficient to solve the much simpler two-component
Eq.(444) and then generate the corresponding Dirac solutions by the use of Eq. (445). In the special
case when the scalar potential A0 (and hence E) vanishes, the two-component equation then has
the canonical form of the Pauli equation describing the motion of a charged particle in an external
magnetic field. If further, m = 0 and the motion is confined to two dimensions, then the Pauli Eq.
(444) exactly reduces to the Eq. (402) of the last section. Further, since
(HD)2=[~.(P+A)]2=H~ali
(446)
354
hence, there is a supersymmetry in the massless Dirac problem in external magnetic fields in two
dimensions since Z-l;, Qi and Q2 (see Eq. (403) satisfy the SUSY algebra as given by Eq. (404).
Clearly, this supersymmetry will also be there in two Euclidean dimensions. We can now immediately borrow all the results of the last section. In particular, it follows that if the total flux @(=
s B, dxdy) =n + E (0 5 E < 1) then there are precisely it - 1 zero modes of the massless Dirac equation
in two dimensions in the background of the external magnetic field B (B E B,) [204,214]. Further,
in view of Eqs. (444) and (445) we can immediately write down the exact solution of the massless
Dirac equation in an external magnetic field in two dimensions in all the four situations discussed in
the last section when the gauge potential depended on only one coordinate (say y). Further using
the results of that section, one can also algebraically obtain the exact solution of the Dirac equation
in an uniform magnetic field in the symmetric gauge when the gauge potential depends on both x
and y.
Even though there is no SUSY, exact solutions of the Pauli and hence the Dirac equation are also
possible in the massive case. On comparing the equations as given by (444) (with A0 = 0, E = 0)
and (402) it is clear that the exact solutions in the massive case are simply obtained from the
massless case by replacing l? by _??- m*. Summarizing, we conclude that the exact solutions of the
massive (as well as the massless) Dirac equation in external magnetic field in two dimensions can be
obtained algebraically in case the magnetic field B(E B,) has any one of the following four forms
(i) B = constant (ii) B = -asech*y (iii) B = -asec*y(-r/2
5 y 2 r/2) (iv) B = -c2exp(-y).
Further, in the uniform magnetic field case, the solution can be obtained either in the asymmetric or
in the symmetric gauge [ 202,lOll.
11.4. SUSY and the Euclidean Dirac operator
In the last few years there has been a renewed interest in understanding the zero modes and the
complete energy spectrum of the square of the Dirac operator for a Euclidean massless fermionic
theory interacting with the background gauge fields [25,28,215,216]. Let us first show that there is
always a supersymmetry in the problem of the Euclidean massless Dirac operator in the background
of the gauge fields. On defining
(447)
where
D, = 8, +iA,,
(448)
(451)
355
This supersymmetry is popularly known as chiral supersymmetry. In view of the above representation
of the y matrices, we find
Y$, =
iDo+;*)
= (;+
;>
(452)
-iDO+o.D
and hence
-H
= (ypDJ2
= D,D,
CT*(B+E)
0
0
cr. (B-E)
>
(L:+SL)
(453)
(454)
Using the techniques of Sections 2 and 4, one can algebraically obtain the eigenvalues and the
eigenfunctions of H for few special cases [ 1011. Further, once we have an eigenfunction of H with
a nonzero eigenvalue El, then one can easily obtain the eigenfunctions C& of ycLD, with eigenvalues
&( El ) I* by the construction
& = (Q,
f (El) *M.
(455)
Finally, it is worth pointing out that in case the gauge potential A, has the special form
A, = f,,&x
(456)
We have seen that there is always a supersymmetry associated with the Euclidean Dirac operator
in four as well as two dimensions in the background of the gauge fields. This SUSY was first
successfully exploited in the context of the study of the chiral anomalies [ 2171. In a nut shell, the
solvability of the Dirac operator is related to the ability to integrate the path integral. The quantities
one wishes to calculate are the Greens functions
G(x, X; 7) = (Xle-H7/X)IX+
(458)
G,(x,x;r)
(459)
and
= (xITr(yge-H)Ix)I,,/.
By using Schwingers proper time formalism we can determine S(x, X; A) from G. Similarly, the
index Zs which in the limit r = 0 is related to the chiral anomaly, is just the spatial integral over Gs.
356
E Cooper
et al. /Physics
267-385
Here we would like to show that by introducing the fermionic degrees of freedom, the path
integral for the Dirac operator, which is initially a path ordered integral gets reduced to an ordinary
path integral. This trick was first introduced by Rajeev [218] who was interested in reformulating
quantum electrodynamics
as a supersymmetric theory of loops. Once we introduce the fermionic
variables then G( G,) is determined by choosing the antiperiodic (periodic) boundary conditions for
the fermions. By integrating over the fermionic degrees of freedom, a purely bosonic path integral is
then obtained. For the case of a constant external field strength F,,, the bosonic path integral is a
Gaussian, which allows one to trivially obtain G and G5.
Let us consider the square of the Dirac operator i.e. (y+uOP). It can also be written as
H = (P + A)2 - $T~F~~
(460)
where
(7PJ =
~[y.Y~l.
(461)
If we do not introduce auxiliary fermions then the related matrix valued Lagrangian
L(x, k) = $k,k,
- iA,i:,
+ ~upF~,,
(xle-HT(x) =
would be
Dx,(r)Pexp[
(462)
s
0
d#L(x,
i)
where P denotes path ordering. Now, taking analogy from SUSY quantum mechanics,
noticed [ 218,101] that one can introduce the Grassman variables +P via
&L = $9
(463)
it has been
(464)
- iA,kp
(465)
now becomes
- itip (&6,,
+ Fp,)~v
(466)
St& = E_$.
(467)
F: Cooper et al/Physics
357
where we impose antiperiodic boundary conditions on the fermion at 0 and 7. Since the fermionic path
integral is quadratic, we can perform the functional integral over the fermionic degrees of freedom
exactly for arbitrary Fpy .
The result of the fermionic path integral is
Det12(i&8,,
+ F,,)
Det//*( i&6,,)
(469)
where the prime denotes the omission of the zero mode. To evaluate the determinant
skew diagonal form:
n
i
tj
cash -
Xid7.
(471)
of
(472)
Thus for any arbitrary external field one can explicitly perform the fermionic path integral and is
left with the purely bosonic path integral which, for example, could by performed numerically by
Monte Carlo techniques. For the particular case of a constant external field one can further explicitly
perform the resulting bosonic path integral, When
A, = -;Fp,x,,
(473)
with Fp,, being a constant matrix, one obtains for the remaining bosonic path integral:
(474)
G and G5 are easy to calculate in two and four Euclidean dimensions for the constant external field
case [ 1011 using this method. In two dimensions, where only Fol = B exists one has that (xi = B)
[loll
G= Btanh(Br);
G5 = B = igp,,Fp,.
(475)
It is worth remarking here that the same result has been obtained with more difficulty by Akhoury
and Comtet [ 161. In four dimensions one has a simpler derivation of the result of Schwinger [ 2191
rewritten in Euclidean space. One finds that
358
x,
= [F +
(F2 - G2)1/2]1/2,
x2
[F -
(F2 - G2)1/2]1/2,
(476)
where
F = ~F,,F,,,
G = $F,,F;,
so that
rI
Xi =
$F,,F;,.
(477)
G(x,x;r)
= F,,F:,n
cash -
2 ,
i=l sinh y
G(x,x;~)
= tF,wF;v.
(478)
Once one has obtained the Greens function it is easy to reconstruct the effective action which
allows one to obtain the rate of pair production from a strong external electric field. The effective
action is given by
x?ff= fixL(x)
L(x)
~ss-~--*~G( s).
(479)
E Cooper et al/Physics
359
12.1. General formalism, negative energy states and breakdown of the degeneracy theorem
Given any nonsingular potential 8(x) with eigenfunctions & (x) and eigenvalues E,( n = 0, 1,2, . . .) ,
let us now enquire how one can find the most general superpotential W(x) which will give v(x)
upto an additive constant [ 1141. To answer this question consider the Schriidinger equation for e(x) :
-cp
+ V(x)#J
= EC/l
(480)
(481)
where we have used Eq. (480) for the last step. The eigenvalues of yCd, are therefore
E nc4)= E, - E.
given by
(482)
One usually takes E to be the ground state energy Eo and C#Jto be the ground state wave function
&J(X), which makes Eocoj = 0 and gives the familiar case of unbroken SUSY. With this choice, the
superpotential W,(x) = -I&/& is nonsingular, since fiO(x) is normalizable and has no zeros. The
partner potential $4) has no eigenstate at zero energy since A&,(x) = [ d/dx + W,(x) ] e,,(x) = 0;
however, the remaining eigenvalues of V&+) are degenerate with those of KC@,.
Let us now consider what happens for other choices of E, both below and above the ground state
energy Eo. For E < EO, the solution 4(x) has no nodes, and has the same sign for the entire range
-oc < x < +oo. The corresponding superpotential W4(x) is nonsingular. Hence the eigenvalue
spectra of VIc4, and I$$, are completely degenerate and the energy eigenvalues are given by Eq.
(482). In particular, Eocdj = E0 - E is positive. Here, W+ has the same sign at z = foe, and we have
the well-studied case of broken SUSY [lo].
If the constant E is chosen in the range Eo < E < El, then the solution 4(x) of Eq. (482) will
have only one node, say at the point x = x,. Near the point x = x,, the node of 4(x) makes the
superpotential W(x) singular
4 = a(x - x,),
4 = a,
VQ, =2(x
- x,)?
[using
(483)
Eq. (481) ]
(484)
It is well-known [221-2231 that for any singular potential V(x) = A(x - x$)-~, the behavior of
the wave function at x = x, is governed by the value of A. For values A > i, the potential V(x) has a
strong singularity which forces @(x, ) = 0 and the range of x is effectively broken into two disjoint
pieces x < x, and x > x, with no communication between them. For -f < A < a, the potential has
a singularity of intermediate strength. It is not strong enough to make $(x,) vanish and in fact the
two regions x < xs and x > x, do communicate and one has the full range -cc < x < +oc. Here, in
360
Table 12.1
Choice of the constant E and the corresponding solution 4(x) in Eq. (480) determines the choice of the superpotential
W, which produces any given non-singular potential P(x) [ Eq. (48 1) 1. The table shows how certain choices of E give
rise to singular superpotentials, negative energy eigenstates and a breakdown of the degeneracy theorem. We have taken
4(x = --03) = 0 for convenience [ 1141.
Choice of
Superpotential
6, 4(x)
W,(x)
Negative energy
states for
Zero energy
state for
Degeneracy of
spectra for V*(4) (x)
Vw, (1)
v,,g, (x)
and b)(x)
yes
[broken SUSY]
En
[ 4 (x) has no nodes]
non-singular
none
no
E = El)
non-singular
none
yes
[ Eoco, =
E <
L@(x) = $0(x)
EC,< E < El
[4(x) has one node]
singular [where
4 (x) has node]
one, at
E = El
singular [where
+I (x) has node]
singular [where
4(x) has nodes]
[@J(x) =+1(x)1
El < E < E2
[ r$( x) has two
nodes]
01
no
no
one, at
Eo(I) = Eo - El
yes
[EI(I, = 01
no [partial degeneracy
for symmetric
potentials]
two, at
no
no
Eo(+p)= Eo -
Eoc+, = Eo - E,
E I(+) = EI - l
principle, one can have singular wave functions which require self-adjoint extensions [ 2211; in what
follows, we will deal with regular solutions. Finally, for h < -i, the Hamiltonian is unbounded from
below.
From Eq. (484)) it is clear that V&b, has a strong singularity which makes J,!I(x,) = 0. Thus, the
problem of finding the eigenstates of V&4, is really two separate problems; one in the range -oo < x,
and the other in the range x, < x < +co. Clearly, this is very different from the range of Vlc~, which
is the whole real axis --oo < x < +co. Hence, in general, the degeneracy theorem obtained from
SUSY for the spectra of V&d, and &cd1 is not valid. The above discussion can be readily extended
to all values of the constant E - the only difference being the number of poles present. A summary
of results is given in Table 12.1.
So far we have considered superpotentials (both nonsingular and singular) which give rise to a
nonsingular potential V,(4). However, if the potential V(x) itself has singularities, then so must all
superpotentials which produce it. Consider the case of a simple pole singularity at x = x0. Then,
near x = x0, the singular superpotential W(x) = g( x - x0) -I gives the following behavior to the
corresponding partner potentials:
K(4)=
g(g+ 1)
(x
xo)2
%Tb)=
dg - 1)
(x
xo)2
Then clearly, for g > 5 or g < -i, both Vi($) and V&4)
two disjoint regions x < x0 and x < x0. Here, one expects
to broken or unbroken SUSY). Similarly, for -i < g < +i,
strength singularities and the whole region --00 < x < +oo
(485)
have strong singularities and both have
degenerate energy levels (corresponding
both V,(4) and V2(41have intermediate
is valid for both. Here again, one obtains
361
degeneracy. However, for the two regions -5 < g < -i and i < g < $ only one of the partner
potentials has a strong singularity, whereas the other has a singularity of intermediate strength.
Therefore, the two potentials have different Hilbert spaces and, in general, degeneracy is not there.
The above discussion is borne out by the work of Jevicki and Rodrigues [ 1 lo] who consider a
superpotential of the form W(X) = g/x - x.
The general discussion of singular superpotentials, negative energy states and breakdown of the
degeneracy theorem is best illustrated with some specific examples.
Consider the harmonic oscillator potential V(X) = x2. The energy eigenvalues are En = 2n and the
first few eigenstates are
(486)
Using the solution $ = +o(x), one gets the usual nonsingular
leading to the following partner potentials and eigenvalues:
=x2+1,
Vl(0) =x2 - 1,
ho,
E$)
n=0,1,2,...
= 2n + 2,
superpotential
W. = -$o//+o
= X,
E;$) = 2n,
(487)
This is standard unbroken SUSY. The harmonic oscillator potential can also be obtained if one starts
from the solution 4 = fiI (x). One gets the singular superpotential WI = -r,bi/#, = x - l/x. It has a
pole at x = 0 and is a special case of the form discussed in Ref. [ 1 lo]. The partner potentials are
Vl(l) = x2 - 3,
v*(,)=x*+--1,
(488)
X2
[54,44].
Ej;;j, =4n+4,
,...
spectra are
(489)
There occurs a negative energy state in Kc ,) at E,$,i)= -2. Th is is expected since we chose 4 to be
the first excited state and E,,(I)1) - 0 , which pushes the ground state to a negative energy. Proceeding
along the same lines, we can get the harmonic oscillator potential using yet another superpotential.
Taking 4 = &(x), one gets W2 = -I+$/& = x - 4x/( 2x2 - 1)) which has poles at x = f l/a.
The
corresponding potentials are
Vl(2) = x2 - 5,
x2_3+
b(2) =
Wx2+l)
(2x2 - 1)2
(490)
The eigenvalue spectrum of K(2) is E,$) = 2n - 4 indicating two negative energy states at -4 and
-2, which was anticipated. The potential V2(2)is not analytically solvable. The strong singularities
< x <
at x = &l/v% break the x-axis into three disjoint regions --oo < x < -lIfi,-l/Jz
l/v&, l/a
< x < +cc and the degeneracy theorem breaks down.
In summary, the potentials vco, and V2(o) give a realization of SUSY QM with a non-degenerate
zero energy ground state and pairing of excited states. For the case of Vl( ,) and V*(r) the eigenvalues
362
Fig. 12.1. The harmonic oscillator potential Vl(1) and its partner potential &(I) as given by Eq. (488). These potentials arise
from a singular superpotential WI = x - l/x corresponding to the choice C$= $1 (x), E = El in Table 12.1 . The energy
levels for both potentials are shown. Notice the negative energy state of VI(~) at energy -2, and the partial degeneracy of
the eigenvalue spectra coming from the even parity of the potentials [ 1141.
are given by Eq. (489). There is partial degeneracy of the spectrum [see Fig, 12.11 due to the fact
that the potentials are symmetric. The states which are missing in Vzclj are the even parity states,
since the 2/x2 barrier requires the vanishing of wave functions at x = 0. Finally, for the partner
potentials V,(2) and V*(2),degeneracy is completely absent.
Our second example is the Morse potential V(x) = A2 + B2evznx - B( a + 2A)e-. The superpotential based on the ground state is W,(x) = A - Bedax. For concreteness, we shall take A = 4, LY= 1
and B = 1. The corresponding energy eigenvalues are E,,C0l
(I) - 16-(4-n)*;
there are four bound states
with eigenvalues 0,7,12 and 15 and a continuum starting above 16. The two lowest eigenfunctions
are [54]
$0(x)
N e
-(4x+C)
$1(x>
we
-(3x+e-)
(7 _ 2e-).
(491)
The supersymmetric partner potentials constructed from the nonsingular superpotential W,(x)
4 - e- are
V1coj = 16 - 9e- + e-,
(492)
The spectrum of h(o) is identical to that of V~(O),except that there is no state at zero energy. An
alternative singular superpotential WI (x) which also yields the Morse potential is
I/J;
wl(x)=-F=
21 - 15e- + 2e-2X
7 - 2e-
(493)
(494)
-3
-2
-1
363
Fig. 12.2. The Morse and its supersymmetric partner potential [eq. (494)] coming from the singular superpotential WI
given in Eq. (493). The singularity at x = - ln3.5 breaks P&I) into two disjoint pieces, and the eigenvalue spectra have no
degeneracy [ 1141.
By construction, the potential Kc i) is the Morse potential and V&l) has a singularly at x, = - ln( 3.5).
Expanding V&1) about the singular point gives V2c1jN 2/(x - x,)~, which requires the wave function
to vanish at x = x,. This effectively breaks the potential and the real axis into two parts: V&i) (left)
for -co < x < x, and V2(ij (right) for x, < x < +oo. The potentials and their eigenstates are
plotted in Fig. 12.2. The potential &(i) has energy levels located at -7, 0, 5, 8 and a continuum
above 9. As expected from Table 12.1, there is a negative energy state at -7. We have calculated the
energy levels of V2(ij (right) and V2c1j (left) numerically. They are Elfi (right) = 6.08, 8.65 with a
continuum above 9 and EC,,
(2) ( left) = 22.96, 50.69, 82.16, 116.8,. . . The potential V2(,) (left) has an
infinite number of bound states and as is obvious from the spectra, the degeneracy between V2(ij and
VlcIj is completely broken by the strong singularity in Wi (x). Note that since the Morse potential
is asymmetric, no partial degeneracy remains. This is unlike our first example (harmonic oscillator)
which had symmetric potentials and V211jand VI( 1j had a partial degeneracy.
Our last example is the reflectionless potential V(x) = A2 - A(A + l)sech2x. The motivation for
considering this example is to make contact with singular superpotentials considered by Casahorran
and Nam [ 1121. The ground state wave function is given by &, = (sechx) A. The nonsingular
superpotential W0 = A tanhx comes from tiO. Using the property of shape invariance [ 54,441 one
readily obtains Q, (x) = tanh x( sechx) A- as the wave function for the first excited state. If one takes
4 = $, , the resulting superpotential is W, = -#I, /I& = -2Acosech2x
+ (A - 1) coth x, which agrees
with Eq. (3.14) of Ref. [ 1121. The energy eigenvalues for vcO, are EL& = A2-(A-n)2
and one has
unbroken SUSY. The potentials generated by Wi (x) are Vlcl) = (A - 1)2 - A( A + 1) sech2x, V&11 =
(A - 1)2 - A(A - l)sech2x + 2cosech2x. These potentials are shape invariant [54,44] and their
(2) - (A - 1) 2 - (A - 2n - 3)2, indicating partial
eigenvalues are E::,, = (A - l)* - (A - IZ)~,E,,(,,
degeneracy, as expected for symmetric potentials. Similarly, our previous experience indicates that
the partner potentials constructed from W,(x) will have no degeneracy. Thus, we have not only
reproduced some families of singular superpotentials previously considered in the literature, but given
the general method to construct new ones.
364
E Cooper
et al./Physics
267-385
In conclusion, we have shown how excited state wave functions can be used to construct singular
super-potentials in SUSY QM. Our method provides a complete and unified picture of the origin
of negative energy states and the presence or absence of degeneracy. Although our technique is
perfectly general, our examples were taken from the class of shape invariant potentials, since these
are analytically solvable.
12.2. Bound states in the continuum
In 1929, Von Neumann and Wigner [ 1161 realized that it was possible to construct potentials
which have quantum mechanical bound states embedded in the classical energy continuum (BICs).
Further developments, by many authors [ 117,224-2261 have produced more examples and a better
understanding of the kind of potential that can have such bound states, although there is not as yet
a fully systematic approach. These authors have also suggested possible applications to atoms and
molecules. Capasso et al. [227] have recently reported direct evidence for BICs by constructing
suitable potentials using semiconductor heterostructures grown by molecular beam epitaxy. Finally, it
is interesting to note that BICs have found their way into a recently written text [ 1181.
In this subsection, we show how one can start from a potential with a continuum of energy
eigenstates, and use the methods of SUSY QM to generate families of potentials with bound states in
the continuum [ BICs] [ 1151. Basically, one is using the technique of generating isospectral potentials
(discussed in Section 7) but this time starting from states in the continuum. The method preserves
the spectrum of the original potential except it adds these discrete BICs at selected energies. As
illustrative examples, we compute and graph potentials which have bound states in the continuum
starting from a null potential representing a free particle and the Coulomb potential.
(a) One parameter family of BICs. Consider any spherically symmetric potential V(r) which
vanishes as r + 00. The radial s-wave Schrodinger equation for the reduced wave function u(r) (in
units where tZ = 2m = 1) is
--u + V(r)u(r)
= l%(r),
(495)
where we have scaled the energy and radial variables such that all quantities are dimensionless. Eq.
(495) has a classical continuum of positive energy solutions which are clearly not normalizable.
As we have seen in Section 7, the Darboux [64] procedure for deleting and then reinstating the
ground state uO( r) of a potential V(r), generates a family of potentials q( r; A) which have the same
eigenvalues as V(r). These isospectral potentials are labeled by a real parameter h in the ranges
A > 0 or h < - 1. The isospectral potential P(r; A) is given in terms of the original potential V(r)
and the original ground state wave function uO(r) by [ 68,70,72]
Q(r;A)
= V(r)
- 2[ln(Z0+
A)]=
4u(&J
21104
V(r) - &+A + (z0+A1*
where
lo(r)
E /ui(r)dr.
J
(497)
E Cooper et al/Physics
365
Except in this section, uo was taken to be the nodeless, normalizable ground state wave function of
the starting potential V(r). However, it is easy to generalize the above equations to the case where
uo(r) is any solution of Eq. (495) with arbitrary energy Eo. If uo (r) has nodes, this leads to singular
superpotentials and to singularities in the partner potential V,(r) . However, when the original state at
EO is re-inserted, the resulting family of potentials p( r; A) is free of singularities [ 1141. Our results
are best summarized in the following statement:
Theorem: Let uo( r) and ul (r) be any two nonsingular solutions of the Schrodinger equation for
the potential V(r) corresponding to arbitrarily selected energies E. and El respectively. Construct a
new potential p( r; A) as prescribed by Eq. (496). Then, the two functions
u0W
iio(r; A) = -
ZOSA
and
a,(r; A) = (-5
- Eo)u, + iioW(uo,ul),
[where W denotes the Wronskian, W( uo, u1) = uoui - u1u;] are solutions of the Schrodinger
for the new potential p( r; A) corresponding to the same energies E. and El.
(499)
equation
While the new potential in Eq. (496) and the new wave functions in Eq. (498) were originally
inspired by SUSY QM, the easiest proof of the above theorem is by direct substitution. One simply
computes -6: + i/( r; A)& (i=O,l), with the wave functions iii given in the theorem. After straightforward but tedious algebraic manipulations, one gets EiGi, thus establishing the theorem. The algebra
is considerably simplified by using the following identity for the Wronskian of two solutions of the
Schrodinger equation:
$Wo>
~1) = (Eo -
4)uoul.
(500)
366
Table 12.2
One and two parameter families of potentials with bound states in the continuum.These families are generated by applying
the theorem described
[ 1151.
Potentials
V(r)
Q = V - 2[ln(Zo + A)]
6=5i-2[ln(il+AI]
El
UI
1
:.
Ul = -&
f, +A1
Eo
uo
1
lie = -ug
lo + A
bo=(Eo-El)iio+&W(a,,Lir~)
Wave functions
to obtain
Example I: Free particle on the half line. Consider a free particle on the half line (V G 0
for 0 2 r < 00). We choose ~0 = sin kr, the spherical wave solution, corresponding to energy
E0 = k2 > 0, which vanishes at Y = 0. The integral I0 given in Eq. (497) becomes
I0 = [2kr - sin(2kr)]/(4k).
(501)
32k2 sin kr
8k2 sin(2kr)
D2
-
Do
(502)
with
DO( r; A) = 2kr - sin(2kr)
+ 4kA.
(503)
(504)
For special values of the parameters k and h, the potential Q and its BIC wave functions are shown
in Figs. 12.3a and 12.3b. The original null potential has now become an oscillatory potential which
asymptotically has a 1 lr envelope. The new wave function at E0 = k2 also has an additional damping
factor of 1 lr which makes it square integrable. As u. appears in the numerator of v, Eq. (496),
every node of 2, is associated with a node of Q but not every node of 9 produces a node of a,. The
value of the eigenenergy E. is clearly above the asymptotic value, zero, of the potential. Evidently,
the many oscillations of this potential, none of them able to hold a bound state, conspire in such a
way as to keep the particle trapped.
The parameter A which appears in the denominator function Do( r; A) plays the role of a damping
distance; its magnitude indicates the value of r at which the monotonically
growing integral 20
becomes a significant damping factor, both for the new potential and for the new wave function. This
1.0
".',"","",""J
ial
r\
I...,,,,
367
,,,,I,/,,,.
ib)
-i
h = 5.0 -1.0
u
0
10
k=l.O
15
20
Fig. 12.3. Potentials P(r) (solid lines) possessing one bound state in the continuum (BE) obtained by starting from a
free particle and constructing a one-parameter (A) family. The BIC wave function iio( r) is at energy E = 1, and is shown
by the dashed line. Fig. (a) corresponds to A = 0.5 and fig. (b) corresponds to a much larger value A = 5 .
is illustrated graphically in Figs. 12.3a and 12.3b which are drawn for very different values of A.
[Note that the wave functions shown in the figures are not normalized]. The parameter A must be
restricted to values greater than zero in order to avoid infinities in 9 and in the wave functions. In
the limit A + co, ? becomes identical to V.
Example 2: Coulomb potential. Here V = Z/Y, and the unbound, reduced 1 = 0 wave function
satisfies the Schrodinger equation Eq. (495), which can be written in standard form
(505)
with p = &!?r and 75= Z/2&.
The solutions involve confluent hypergeometric functions which in the asymptotic limit approach
sine waves phase-shifted by a logarithmic term. Useful expressions for these solutions in the regions
near and far from the origin are available in the literature [ 228,229]. Stillinger and Herrick [ 1171,
following the method of Von Neumann and Wigner [ 1161, have constructed BIC potentials and wave
functions for the case of the repulsive Coulomb potential. Here we use our theorem to construct a
one-parameter family of isospectral potentials containing a BIC. The procedure is the same for both
positive and negative Z; the only difference being in the sign of q. The formal expressions for the
BIC potentials and wave functions have been given above, Eqs. (496) and (498), in terms of uo.
The positive energy solution of Eq. (505) can be written in the usual form [ 228,229,117] as the
real function
u0(p>
= C0(7?>e
+M(
1 - i*, 2,2ip),
(506)
+ifj)(
(507)
where
Co(q) = (e-2)]r(1
and M(a, b; z) is Kummers function. Using tabulated expressions for the Coulomb wave functions
[229] and doing the integral for lo numerically, we have obtained the BIC wave functions for
representative values of A. The corresponding one-parameter family of potentials obtained by the
SUSY procedure is given in Eq. (496) with V, = Z/r.
368
(b)
2 = -2
k = 0.5
A = 0.25
Fig. 12.4. (a) A potential (solid line) with one BIC at energy E = 0.25 obtained by starting from an attractive Coulomb
potential (dotted line). (b) A plot of the wave functions corresponding to the potentials in part (a).
The results are displayed in Fig. 12.4. Fig. 12.4a shows the BIC partner to the attractive Coulomb
wave function of the
potential for A = 1, k = 1, and 2 = -2. Fig. 12.4b shows the (unnormalized)
bound state in the continuum for this potential at E. = k*. For comparison the original Coulomb
potential and wave function are also shown dotted. It is seen that the potential which holds a bound
state of positive energy shows an oscillatory
behavior about the Coulomb potential, V,, as is also
.
evident from the form of Eq. (496) for V. Since the oscillating component vanishes whenever u.
vanishes, we have V = V at each node of ~0. Compared to the original, unnormalizable wave function,
the BIC wave function in both cases shows a damped behavior due to the denominator function. This
is also seen in the figures.
A similar behavior is also expected for other spherically symmetric potentials with a continuous
spectrum of positive eigenvalues. For arbitrary one-dimensional potentials, where the range extends
from --oo to +oo, the situation is not so clear cut. Our method works for the Morse potential which
is steeply rising on the negative x-axis with correspondingly damped wave functions. It also works
for the case of a particle in a constant electric field for similar reasons. For potentials, such as
V(x) = -Vosech2x, the integral 10 in Eq. (497) is not convergent if the starting point is chosen at
-co, and it gets negative contributions if the starting point is selected at finite x-values. This leads
to a vanishing denominator function in the expressions for some wave functions which makes them
unacceptable.
(b) Two parameterfamily ofpotentiaZs. In the previous section (a), we have seen how to generate
a one-parameter family of potentials with one BIC. We now show how this procedure can be extended
to construct two-parameter families which contain two BICs.
In constructing the new wave functions for the one-parameter family, Eq. (496), we observe that
the denominator function given in Eq. (498) was all that was needed to create the BIC, while the
operation in Eq. (499) ensured that the wave functions for all the other states, there represented by
Gl, are a solution to the new potential. Note again, there is nothing special about the ordering of the
two energy values nor the relative magnitude of E. and El, therefore we can repeat this procedure
by applying the theorem to the wave functions and the potential of the one-parameter family, but this
time we transform the state at El into a BIC. The state at Eo, which already is a BIC, is transformed
in the step of Eq. (499), suitably modified, to become a solution to the new potential. In this way
we obtain the two parameter family of potentials
369
(508)
with the solutions of the corresponding
Schrijdinger
equation
~o=(Eo-E*)iio+a,w(ii,,a,),
2
(509)
a1
(510)
Ul
=jl$-
and
il -
#(r)dr.
(511)
The precise relationship of the new potential and its wave functions, which are now both BICs, is
illustrated in the last column of Table 12.2.
While the compact form of Eqs. (508)-( 510) explicitly shows the method of construction, it is
useful to observe that the integral & can be conveniently re-cast into a simpler form which contains
integrals of the form
r
Zi =
uf(r)dr,
(512)
u;w
involving the original wave functions only. Making use of Eq. (499) for B1, we get
J[6%
r
i, =
Eo>~u:
cIo+Aj2
+~(EI
-Eo)
cI~~$W
(513)
dr'.
J
0
W2(r)dr
(IO + A)2
-I@ r+
= lo+Ao I/
2?Vw dr,
(514)
(Jo+A)
We now use Eq. (500) for the derivative of a Wronskian of two solutions of the Schrodinger equation
to rewrite the second term and observe, that it exactly cancels the last term in Eq. (513). We therefore
have
f,(r)
--WV-)
= I + A + (El - Eo)2Mr).
(515)
Here we have made use of the fact that our boundary conditions
As an example, we evaluate the two-parameter potential
(Zo+A)[(E,
-Eo)211
= 0.
(516)
(517)
370
Fig. 12.5. (a) An example of a potential with two BICs at energies EO= 1 and El = 4. (b) The wave functions at energies
Eo=l (dashedline) andEl =4 (dottedline) [115].
We happen to
step, the state at
procedure in the
then the state at
corresponding to
have transformed first the state at energy E. into a BIC and then, in the second
El, which introduced the parameters h and A,. Let us now consider applying our
reverse order, that is let us first transform the state at energy El into a BIC and
energy Eo, producing the parameters ,u and pI. For this situation, the argument
Eq. (517) is
if the parameters
(518)
are related by
(519)
This also leads to the same two-parameter wave functions. We also note that transforming any state
twice by Eq. (498) does not create a second denominator or anything else new, but simply changes
the value of the parameter A as shown in Ref. [70]. Finally, relation (509) ensures that all other
eigenstates will be solutions to the new potentials.
Shown in Fig. 12.5 is a potential with two BICs at energies Eo = 1, El = 4. Clearly, the above
procedure can be readily extended to obtain multi-parameter families with multiple BICs at arbitrarily
selected energies.
Our discussion of BICs has been restricted to effectively one dimensional problems, and as stated
before there is now some experimental evidence for the existence of BICs under appropriately chosen
conditions [ 2271. Recently, there has also been a computation which claims the existence of BICs
in QED in three dimensions [ 2301. These authors have also speculated that the BIC energies they
compute in e+e- scattering in QED are in fact in reasonable agreement with unusual peaks observed
in recent heavy ion experiments [ 23 11.
13. Parasupersymmetric
quantum
mechanics
and beyond
In the last few years, exotic quantum statistics have been widely discussed in the literature. For
example, in two space dimensions, one can have a one-parameter family of statistics interpolating
371
between Bose and Fermi statistics [232]. On the other hand, in three and higher space dimensions
parafermi and parabose statistics [233-2361 are the natural extensions of the usual Fermi and Bose
statistics. In particular, whereas Fermi and Bose statistics describe the two one-dimensional representations of the permutation group, parafermi and parabose statistics describe the higher dimensional
representations of the same group. In view of the fact that the SUSY has provided us with an elegant
symmetry between fermions and bosons, it is natural to enquire if there exists a generalization which
includes the above exotic statistics. Such a question was raised many years ago in the context of
parastring models [237], but the specific symmetry algebra was still the usual SUSY one.
In this section, we study the possibility of having a symmetry between bosons and parafermions.
We shall construct parasupersymmetric
quantum mechanics (PSQM) of a boson and a parafermion
of orderp(=
1,2,3,. . .). It turns out that whereas in the usual SUSY QM, the symmetry generators
obey structure relations that involve bilinear products, in PSQM of order p, the structure relations
involve products of (p + 1) parasupersymmetry
(PARASUSY) charges. Various consequences of this
algebra are also discussed.
It is worth adding here that historically the PSQM of order 2 was introduced first [ 1061 and its
various consequences were discussed [ 238-2421. Initially it was felt that the generalization to order
p was not possible in the sense that the PSQM of order p cannot be characterized with one universal
algebraic relation [ 2431. However, later on, one was indeed able to construct such a PSQM of order
P
[1071.
Very recently, new forms of quantum statistics called orthofermi and orthobose statistics, have been
constructed [ 2441. It is then natural to construct orthosupersymmetric
quantum mechanics (OSQM)
where there is symmetry between a boson and an orthofermion of order p [ 1091. Unlike the PSQM
of order p, one finds that the structure relations now involve only bilinear products of the symmetry
generators. Various consequences of SUSY QM, PSQM and OSQM are also discussed in this section.
In particular, it is worth pointing out that whereas in SUSY QM and OSQM of order p, the energy
eigenvalues are necessarily nonnegative, in PSQM of order p they need not be so.
13.1. Parasupersymmetric
quantum mechanics
In order to motivate the algebra of PSQM, let us recall that in SUSY QM the symmetry
a boson and a fermion is characterized by the algebra
Q=O=Q+*,
[f&Q]
=O,
QQ++Q+Q=2H.
between
(520)
Note that there is an extra factor of 2 on the right hand side compared to the algebra given in Section
2. This results from assuming m = 1 in this section, in conformity with the notation followed by
the various authors in this field. The SUSY QM algebra is easily motivated by recalling that the
fermionic operators a, a+ satisfy the algebra
a2=O=at2,
{a,a+}=l.
A useful representation
a=(:
h),
(521)
i).
(522)
372
E Cooper et al./Physics
operators
(b)P+ = 0 = (bt)P+,
(523)
=-2b,
[[&b],b]
[[b+,b],bt]
=2bt.
(524)
On letting
J+ =b+,
J- =b,
it immediately
algebra
[J+,J_]
J3 = ;[bt,b],
[J3,Jk]
(525)
and (524) that the operators
=%J*.
(526)
c,g-l,...,
-;+l,-$
x (p+l)
of the W(2)
(527)
It is now easily
(p+l)
by the following
(528)
= Cp&+~,p
where
c, = &(P
(52%
- P + 1) = cp-PSI.
It is easily checked that the operators b and b+ indeed satisfy the algebra as given by Eqs. (523)
and (524). One can now ask as to what multilinear relation is satisfied by b and b+ apart from the
one given by Eq. (523)? It turns out that the nontrivial relation is
bPb+ +bP-lbtb+...+
l)(p
+2)&-l,
(530)
where one has (p + 1) terms on the left hand side. As expected, for p = 1 this reduces to the bilinear
relation for the fermionic operators given in Eq. (521) .
This relation between b and b+ strongly suggests that one may have an analogous multilinear
relation in the algebra of PSQM of order p. To that purpose, let us choose the PARASUSY charges
Q, and Q/ as (p + 1) x (p + 1) matrices as given by
G () = (Q,t)p+.
= JW,,~
where (r=1,2,...,p)
(531)
satisfy
(532)
(ti = m = 1)
(533)
E Cooper et al./Physics
H Pfl
commutes
wf_,
= ;
;(w; +
w;>+
373
;cp
(534)
(s = 2,3, . . . , p)
(535)
= wf - w, + c,.
Here C1, C,, . . . , C, are arbitrary constants with the dimension of energy. It turns out that the nontrivial
relation between Qi, Q/ and H is given by [ 1071
Qf Qf + Q;-Q~QI+.
and Hermitian-conjugated
(536)
provided
c, + c, + * * . + c, = 0.
(537)
An example is in order at this stage to illustrate the structure of PSQM of order p. If one chooses
w, = w, = . . . = w, = 0)x,
(538)
(539)
(533) takes a very simple form given by
- J30 3
(540)
where J3 is as given by Eq. (527). This H describes the motion of a particle with spin p/2
oscillator potential and a uniform magnetic field. The spectrum of this Hamiltonian is
E,,,, = (n + ; - m)w,
in an
(541)
H, - ;Cr
0
0
Hr+l-;C,
>
(542)
374
(iii)
E Cooper et al./Physics
Apart from Ql, there are (p- 1) other conserved, independent PARASUSY charges Q2, Q3, . . . , Q,,
defined by (s=2,3,...,p)
Qs = (J - iwpPo,p+~
= -(P
- iw$G,p+,
if P$s,
if p = s,
(543)
all of which commute with the Hamiltonian (534) and also satisfy the PSQM algebra as given
by Eqs. (532) and (536) [107].
(iv) There is an interesting application of PSQM in nonrelativistic quantum mechanics. As we have
seen in Section 3, given any Hamiltonian Hi with p bound states with energies El, EZ, . . . , Ep
and the corresponding eigenfunctions +9,, $2, . . . , t,b,,, one can always generate p other Hamiltowith the same spectrum as Hi except that 1,2, . . . , p levels respectively
nians J32, H3,. . . , HQ+I)
are missing from them. Further in that case there are p supersymmetries with the corresponding
Hamiltonians being precisely given by Eq. (542). Besides, the p constants Cl, C,, . . . , C,, are
related to the energy eigenvalues by (r = 1,2, . . . , p)
C,=%EpfE,,,+
P
+. . + E,+, - (P-l)E,+E,-,+...+E,].
(544)
Thus instead of associating the symmetry algebra sE( l/l) @ SU(2) or U( 1) @ SU(2) as we
did in Section 3, one can also associate parasupersymmetry
of order p to the hierarchy of
Hamiltonians H,, Hz, . . . , HP [ 1071.
(v) What is the most general solution of the relation (535) which must be satisfied in order to
have the PSQM of order p? Treating W, f W,._, as the two variables, Eq. (535) can be reduced
to a simple nonlinear equation which can be immediately solved [ 1061. This is however not
very useful as it gives us W, + W,._, for a given W, - W,._,. Ideally, we would like to know
the most general solution for W2 for a given WI and then using this W,, one would recursively
obtain W3, W4, . . . , W,. Unfortunately, this problem is still unsolved. Of course shape invariant
potentials satisfy Eq. (535) but shape invariance is clearly not necessary. As discussed above,
given any H with p bound states, one can always construct WI, W2, . . . , W, which will satisfy
Eq. (535).
One unsatisfactory
feature of the PSQM of order p is that except for p = 1, H cannot be
directly expressed in terms of the PARASUSY charges Q, and Qt. This is because, in Eq. (536))
H is multiplied by Qy- whose inverse does not exist. Another unsatisfactory feature is that unlike
in SUSY QM, the energy eigenvalues are not necessarily nonnegative and there is no connection
between the nonzero (zero) ground state energy and the broken (unbroken) PARASUSY. It turns
out that in case all the constants C1, C2, . . . , C, are chosen to the zero, then one can take care of
both of these unsatisfactory features [ 1081. In particular, in that case H as given by Eq. (534) can
be expressed in terms of any one of the p PARASUSY charges Qs by (s = 1,2, . . . , p )
H = $Q:Qs
- QsQ,t>'+Q,'QsQsQfl"'~
(545)
It is immediately clear from here that all the energy eigenvalues are necessarily nonnegative and
that the ground state energy being 0 (> 0) corresponds to unbroken (broken) PARASUSY. Further,
one can also show that in this case all the excited states are always (p + 1)-fold degenerate while
375
the nature of the ground state will depend on the specific form of W, [ 1081. In the limit of p --f M
it has been shown that the relation (545) reduces to
H = ;QJ&
(546)
which can be termed as the PSQM of infinite order whose algebra corresponds to that of Greenbergs
infinite statistics [ 2451.
There is an interesting application of this version of PSQM to the strictly isospectral Hamiltonians
discussed in Section 7. In particular, one can show that any potential V, = q - Wi with at least one
bound state, forms PSQM of order p along with its SUSY partner potential V, = q + W[ and with
the strictly isospectral potential families V,( X, AZ), V, (x, A3), . . . , V,( x, AP) where AZ, As, . . . , A, are
arbitrary parameters which are either > 0 or < - 1 [ 2031.
Eq. (545) shows an explicit expression for the Hamiltonian H which involves just one charge Qs
but an overall square root. There is an alternative expression for H where it is not necessary to take
the square root, but which involves all the p PARASUSY charges [ 1081:
2H = QrQ!+ Q!Qri- i
where, r,s=
&QjQs
s=l
+ Q:Qr -
2Q;Q,.)
(547)
1,2,.
QrQ,tQr = 2QrH.
In case one chooses W, to be of form ((w = 1,2,. . .,p)
W
a
=_A+a-1
x
then one has a model for conformally invariant PSQM of order p. In this case one can show that the
dilatation operator D and the conformal operator K defined by
D=-$(xP+Px),
z+
(550)
satisfy relations analogous to those given in Eqs. (545)) (547) and (536) in terms of the PARASUSY
charges Qn and parasuperconformal
charges S, defined by [a = 1,2, . . . , p; i, j = 1,2, . . . , (p + 1) ;
r=2,3,...,p]
(551)
(Sl)ij
-x4+1,j,
(Sr)ij
-XSi+l,j
if i $r + 1,
Xsi+l,j
if i=r+
13.2. Orthosupersymmetric
1.
(552)
quantum mechanics
Recently, Mishra and Rajasekaran [244] have introduced new forms of quantum statistics called
orthofermi and orthobose statistics. Orthofermi statistics contain a new exclusion principle which is
more stringent than the Pauli exclusion principle: an orbital state shall not contain more than one
particle, whatever be the spin direction. The wave function is thus antisymmetric in spatial indices
376
alone with the order of the spin indices frozen. In an analogous way, one can also define orthobose
statistics. All these properties follow provided the corresponding creation and annihilation operators
Ct and C satisfy
ckc&@
* 6-+2 c,!,gam,,,$ky = ~krn&+
(553)
ck&n,
(554)
y=l
(h&k,
= 0,
where the upper and the lower signs lead to the orthofermi and the orthobose cases respectively and
the Latin indices k, m, . . . and the Greek indices (Y,j3, y, . . . correspond to space and spin indices
respectively.
For constructing the quantum mechanics of a boson and an orthofermion, we ignore the spatial
indices in Eqs. (553) and (554) and obtain
C&J + s,,-&$,
= a,,,
y=l
C&=0.
Eq. (555)
(555)
(556)
implies that
c,c,t=c~c~=...=cPcpt.
(557)
Following the discussion of the last subsection it is easy to see that a useful representation of
these operators is in terms of the (p + 1) x (p + 1) matrices defined by (cr, p = 1,2, . . . , p; r, s =
132,.
--,(P+l))
(Ca>m =
h&,a+1;
w;)rs
= &,l&,a+l*
(558)
Let us now try to write down the algebra for the OSQM of order p where there is a symmetry
between a boson and an orthofermion of order p [ 1091. On comparing the algebra for the fermionic
and the orthofermionic
operators as given by Eqs. (521), (555) and (556) and remembering the
SUSY QM algebra as given by Eq. (520) it is easy to convince oneself that the p ORTHOSUSY
charges Qa, Qd and the Hamiltonian must satisfy the algebra ((Y, /3 = 1,2, . . . , p )
QaQ;+s,,eQ;Qy ='W,H,
(559)
y=l
QaQp=O, [XQal=O.
(560)
Note that for p = 1 we recover the usual SUSY QM algebra. It is easily checked that if we choose
the p ORTHOSUSY charges Qa as (p + 1) x (p + 1) matrices as given by (r = 1,2, . . . , (p + 1) )
(f'-iWa)&,,&,,+,;
(Q:>rs
= (f'
+iWa)%1&,,+1
(Qa)rs=
and the Hamiltonian
(H),,
where (r=
= H&,,
1,2,...,p)
(561)
H as
(562)
377
then the OSQM algebra as given by Eqs. (559) and (560) is indeed satisfied provided
w;+w;=w;+w;.
Note that this condition
(564)
(565)
which follow from the OSQM relation (559). Thus, unlike PSQM, the constants C1, C,, . . . are not
allowed in OSQM. The various consequences of the OSQM have been discussed in detail in [ 1091.
At this point, it may be worthwhile to make a relative comparison of SUSY QM with PSQM and
OSQM.
(i) First of all, the close similarity in structure between OSQM and PSQM must be noted. For order
p, both are based on (p + 1) x (p + 1) matrices and the structure of H in the two cases is very
similar. The chief difference between the two is the absence of the constants C1, C,, . . . , C,, in
the former while in the latter they may or may not be zero.
(ii) Whereas the ground state energy is zero (> 0) in case the SUSY or the ORTHOSUSY is
unbroken (spontaneously broken), in general, there is no such restriction in the PSQM case and
the energy eigenvalues can even be negative. However, in the special case when C1, C,, . . . , C,
are all zero then the PSQM has similar prediction as the other two.
(iii) Whereas in the SUSY QM, all the excited states are necessarily two-fold degenerate, in OSQM
of order p they are necessarily (p + 1 )-fold degenerate. On the other hand, in the PSQM of
order p, all the levels starting from the Pth excited state (and above) are necessarily (p + l)fold degenerate except when all the constants are zero in which case all the excited states are
necessarily (p + 1 )-fold degenerate.
14. Omitted
topics
So much work has been done in the area of SUSY QM in the last 12 years that it is almost
impossible to cover all the topics in such a review. We have therefore decided to give a brief
description of some of the omitted topics. For each topic, a few references are provided, so that
interested readers can go back and trace other references and get a good idea of the developments.
(i) Supermathematics
[ 2461.
(ii) SUSY in atomic physics [ 41,421. In a series of papers, Kostelecky et al. [41] have discussed
the relationship between the physical spectra of different atoms and ions using SUSY QM. In
particular, they have suggested that the helium and hydrogen spectra come from SUSY partner
potentials. This connection has been commented upon by Rau [42].
(iii) SUSY in condensed matter and statistical physics [ 163,248-2521.
Ideas from SUSY have
been used to give insight into random magnetic fields in Ising-like models, polymers, electron
localization in disordered media and ferromagnets. These topics are covered in a set of review
talks found in the Proceedings of the Conference on Supersymmetry held at CNLS in 1983
[151.
318
(iv)
QWH)
(x)
(xi)
H'"+"Q(a,),
(566)
one can derive recursion relations for the propagators of the hierarchy. If one member the
hierarchy has a known propagator (such as the harmonic oscillator or the free particle) then
it is easy to use these recursion relations to derive the propagators for the heirarchy. This also
allows one to calculate path integrals for SIP.
SUSY and N-body problem [260,150]. A novel algebraic structure is found for l/r* family
of many body problems which provide a novel link between SUSY and quantum integrability
[ 2601. Also of considerable interest is a supersymmetric generalization of a N-particle quantum
mechanical model with combined harmonic and repulsive forces [ 1501.
Coherent states for SIP [ 2611. Recently a Lie algebraic treatment of the SIP has been given
and using it coherent states have been constructed for these potentials.
(xii)
(xiii)
(xiv)
(xv)
(xvi)
379
SUSY and new soliton solutions [ 2621. It is well known that the stability equation for the kink
solutions in scalar field theories in 1 + 1 dimensions is a Schrodinger-like equation. It has been
shown that new scalar field theories with kink solutions can be obtained by considering the
isospectral deformation of this problem.
SWKB and tunneling [ 263,821. Expression for transmission coefficient T through potential
barriers has been obtained within the SWKB approximation and it has been shown that the
analytic continuation of T for the inverted SIP (with translation) leads to the exact bound state
spectrum.
Time-dependent
Pauli equation [ 2641. Kostelecky et al. [264] have succeeded in factorizing
the time dependent Pauli equation and utilizing the SUSY to solve the Pauli equation for a time
dependent spatially uniform magnetic induction.
SUSY and nuclear physics [ 247,157,265].
Relations between the spectra of even-even and
neighbouring even-odd nuclei have been obtained by using the concepts of SUSY [ 2471. Also,
Baye [ 1571 has shown that the deep and shallow nucleus-nucleus potentials which have been
successfully used in the past are in fact SUSY partner potentials. It has also been suggested
that the relationships seen between the energy levels of adjacent superdefotmed nuclei can be
understood in terms of SUSY [265]
q-supersymmetric
quantum mechanics. Inspired by the recent advances regarding quantum
groups and algebras, Spiridonov [61,62] has proposed a deformation of the SUSYQM. In
particular, he has explicitly defined q-SUSY algebra and provided its explicit realization on the
Hilbert space of square integrable functions. It is worth noting that in this case the supercharges
are not conserved as they do not commute with the Hamiltonian. However, in the limit of q
going to 1, one recovers the conventional SUSY QM. Further, in this approach the spectrum of
HZ and H1 are not degenerate but rather the spectrum of Hz is obtained from that of HI just
by q2-scaling i.e. Ec2) = q2E().P ossible exception concerns only the lowest level exactly as in
ordinary SUSY QM. In particular, if the q-SUSY is unbroken (spontaneously broken) then the
ground state energy of HI is zero (> 0).
Acknowledgments
It is a pleasure to thank all of our many collaborators without whose efforts this work would
not have been accomplished. We would especially like to thank Joe Ginocchio for carefully reading
the manuscript. We would like to thank Los Alamos National Laboratory for its hospitality during
the writing of the manuscript. We gratefully acknowledge the U.S. Department of Energy for partial
support.
References
[ l]
[2]
[ 31
[4]
[5]
1.
380
[6] For more details and other references in this field see, for example, M.F. Sohnius, Phys. Rep. 128 ( 1985) 39.
[56]
[57]
[58]
[59]
[60]
[61]
[62]
[63]
[64]
[ 651
[66]
[ 671
[68]
[69]
[70]
[71]
[72]
[73]
[74]
[75]
[76]
[77]
[78]
[79]
[80]
[ 8 1]
[82]
[ 831
[84]
[85]
[ 861
[87]
[88]
[89]
[90]
[91]
[ 921
[93]
[94]
[95]
[96]
[97]
[98]
[99]
[ 1001
[ 1011
[ 1021
[ 1031
[ 1041
[ 1051
381
G.A. Natanzon, Vestnik Leningrad Univ. 10 (1971) 22, Teoret. Mat. Fiz. 38 (1979) 146.
J.N. Ginocchio, Ann. Phys. 152 (1984) 203; ibid. 159 (1985) 467.
A. Khare and U. Sukhatme, Jour. Phys. A 26 (1993) L901.
D. Barclay, R. Dutt, A. Gangopadhyaya, A. Khare, A. Pagnamenta and U. Sukhatme, Phys. Rev. A 48 ( 1993) 2786.
A. Shabat, Inv. Prob. 8 (1992) 303.
V. Spiridonov, Phys. Rev. Lett. 69 ( 1992) 298.
V. Spiridonov, Mod.Phys. Lett. 7 ( 1992) 1241.
A. Degasperis and A. Shabat, Univ. de Roma preprint No. 919 (1992)
G. Darboux, C.R. Academy SC. (Paris) 94 (1882) 1456.
P Abraham and H. Moses, Phys. Rev. A 22 ( 1980) 1333.
D. Pursey, Phys. Rev. D 33 (1986) 1098, 1103,2267; M. Luban and D. Pursey, ibid. D 33 (1986) 431.
K. Chadan and PC. Sabatier, Inverse Problems in Quantum Scattering Theory, Springer Verlag ( 1977).
M.M. Nieto, Phys. Lett. B 145 (1984) 208.
C.V. Sukumar, Jour. Phys. A 18 ( 1985) 2917,2937.
A. Khare and U. Sukhatme, Jour. Phys. A 22 (1989) 2847.
A. Khare and U. Sukhatme, Phys. Rev. A 40 (1989) 6185.
W.-Y. Keung, U. Sukhatme, Q. Wang and T. Imbo, Jour. Phys. A 22 ( 1989) 987.
W. Kwong and J. Rosner, Prog. Theor. Phys. Supp. 86 (1986) 366.
C.V. Sukumar, Jour. Phys. A 19 (1986) 2297; ibid. A 20 (1987) 2461.
Q. Wang, U. Sukhatme, W.-Y. Keung and T. Imbo, Mod. Phys. Lett. A 5 (1990) 525.
W. Kwong, H. Riggs, J. Rosner and H.B. Thacker, Phys. Rev. D 39 (1989) 1242.
T. Imbo and U. Sukhatme, Phys. Rev. Lett. 54 (1985) 2184.
F. Cooper and P. Roy, Phys. Lett. A 143 ( 1990) 202.
A. Comtet, A. Bandrauk and D.K. Campbell, Phys. Lett. B 150 (1985) 159.
A. Khare, Phys. Lett. B 161 (1985) 131.
For a good review on this field and references therein see, R. Dutt, A. Khare and U. Sukhatme, Am. Jour. Phys. 59
(1991) 723.
R, Dutt, R. De, R. Adhikari and A. Comtet, Phys. Len. A 152 (1991) 381.
R. Adhikari, R. Dutt, A. Khare and U. Sukhatme, Phys. Rev. A 38 ( 1986) 1679.
K. Raghunathan, M. Seetharaman and S. Vasan, Phys. Lett. B 188 ( 1987) 351.
D. Barclay and C.J. Maxwell, Phys. Lett. A 157 (1991) 351.
R. Dutt, A. Khare and Y.P. Varshni, Phys. Lett. A 123 (1987) 375.
P. Roy, R. Roychoudhury and Y.P. Varshni, Jour. Phys. A 21 (1988) 1589.
D. DeLaney and M.M. Nieto, Phys. Lett. B 247 (1990) 301.
Y.P. Varshni, Jour. Phys. A 25 (1992) 5761.
A. Khare and Y.P. Varshni, Phys. Lett. A 142 (1989) 1.
S. Fricke, A.B. Balantekin, P. Hatchell and T. Uzer, Phys. Rev. A 37 (1988) 1686.
Y. Murayama, Phys. Lett. A 136 ( 1989) 455.
U. Sukhatme and A. Pagnamenta, Phys. Lett. A 151 (1990) 7.
A. Inomata and G. Junker, Lectures on Path Integration, editors: H. Cerdeira et al, World Scientific, Singapore
( 1992) ; Symposium on Adv. Topics in Quantum Mechanics, editors: J.Q. Liang et al, Science Press, Beijing ( 1992).
R. Dutt, A. Gangopadhyaya, A. Khare, A. Pagnamenta and U. Sukhatme, Phys. Lett. A 174 ( 1993) 363.
R. Dutt, A. Gangopadhyaya, A. Khare, A. Pagnamenta and U. Sukhatme, Phys. Rev. A 48 (1993) 1845.
A. Inomata, G. Junker and A. Suparmi, Jour. Phys. A 26 (1993) 2261.
A. Das and W. Huang, Phys. Rev. D 41 (1990) 3241.
A. Jayannavar and A. Khare, Phys. Rev. D 47 ( 1993) 4796.
B. Thaller, The Dirac Equation, Springer Verlag (1992).
F. Cooper, A. Khare, R. Musto, and A. Wipf, Ann. Phys. 187 ( 1988) 1.
Y. Nogami and F. Toyama, Phys. Rev. A 47 (1993) 1708.
C.V. Sukumar, Jour. Phys. A 18 (1985) L697.
E. DHoker and L. Vinet, Phys. Lett. B 137 ( 1984) 72.
H. Yamagishi, Phys. Rev. D 29 ( 1984) 2975.
382
[ 1071 A. Khare, Jour. Phys. A 25 (1992) L749, Jour. Math. Phys. 34 (1993) 1277.
[ 1081 A. Khare, A.K. Mishra and G. Rajasekaran, Mod. Phys. Lett. A 8 (1993) 107.
[ 1091 A. Khare, A.K. Mishra and G. Rajasekaran, Int. Jour. Mod. Phys. A 8 (1993) 1245.
[ 1lo] A. Jevicki and J. Rodrigues, Phys. Len. B 146 (1989) 55.
[ 1111 M. Shifman, A. Smilga and A. Vainshtein, Nucl. Phys. B 249 ( 1988) 79.
[ 1121 J. Casahorran and S. Nam, Int. Jour. Mod. Phys. A 6 (1991) 2729; J. Casahorran, Phys. Lett. B 156 (1991) 1925:
P. Roy, R.Roychoudhury and Y.P. Varshni, Jour. Phys. A 21 (1988) 3673.
[ 1131 P. Roy and R. Roychoudhury, Phys. Rev. D 32 ( 1985) 1597.
[ 1141 P. Panigrahi and U. Sukhatme, Phys. Lett. A 178 (1993) 251.
[ 1151 J. Pappademos, U. Sukhatme and A. Pagnamenta, Phys. Rev. A 48 ( 1993) 3525.
[ 1161 J. von Neumann and E. Wigner, Zeit. Phys. 30 ( 1929) 465.
[ 1171 F. Stillinger and D. Herrick, Phys. Rev. A 11 (1975) 446.
[ 1181 L.E. Ballentine, Quantum Mechanics, Prentice Hall, PP. 205 ( 1990).
[ 1191 L.E. Gendenshtein and IV. Krive, Sov. Phys. Usp. 28 (1985) 695.
[ 1201 R. Haymaker and A.R.P. Rau, Am. Jour. Phys. 54 (1986) 928.
[ 1211 D. Lancaster, Nuovo Cimento A 79 (1984) 28.
[ 1221 G. Stedman, Euro. Jour. Phys. 6 (1985) 225.
[ 1231 A. Lahiri, P. Roy and B. Bagchi, Int. Jour. Mod. Phys. A 5 (1990) 1383.
[ 1241 0. de Lange and R. Raab, Operator Methods in Quantum Mechanics, Oxford Univ. Press ( 1991).
[ 1251 A. Khare and U. Sukhatme, Physics News (India) 22 (1991) 35.
[ 1261 L. Landau and E. Lifshitz, Nonrelativistic Quantum Mechanics 3rd Volume, Pergamon Press, N.Y. ( 1977).
[ 1271 G. Junker, Jour. Phys. A 23 (1990) L881.
[ 1281 R. De, R. Dutt and U. Sukhatme, Jour. Phys. A 25 (1992) L843.
[ 1291 A. Gangopadhyaya, P. Panigrahi and U. Sukhatme, Helv. Phys. Acta 67 ( 1994) 363.
[ 1301 V.A. Kostelecky, M.M. Nieto, D.Truax, Phys. Rev. D 32 ( 1985) 2627.
[ 1311 A. Khare and R.K. Bhaduri, Inst. of Physics, Bhubaneswar preprint IP-BBSR/93-65; hep-th19310104.
[ 1321 A. Khare and R.K. Bhaduri, Jour. Phys. A 27 (1994) 2213.
[ 1331 M.M. Nieto, Am. Jour. Phys. 47 (1979) 1067; M.M. Nieto and L.M. Simmons,ibid. 47 (1979) 634.
[ 1341 R. Montemayor and L.D. Salem, Phys. Rev. A 40 (1989) 2170
[ 1351 V. Spiridonov, Mod. Phys. Lett. A 7 (1992) 1241.
[ 1361 L.C. Biedenham, Jour. Phys. A 22 ( 1989) L873.
[ 1371 A.M. Macfarlane, Jour. Phys. A 22 ( 1989) 4581.
[ 1381 G.L. Lamb, Elements of Soliton Theory, Wiley, N.Y. (1980).
[ 1391 A. Das, Integrable Models, World Scientific, Singapore ( 1989).
[ 1401 See for example, P.M. Morse and H. Feshbach, Methods of Theoretical Physics, Vol.1, Chapter 5, McGraw Hill, N.Y.
(1953).
[ 1411 F. Calogero, Jour. Math. Phys. 10 (1969) 2191.
[ 1421 J. Wolfes, Jour. Math. Phys. 15 (1974) 1420.
[ 1431 B. Sutherland, Jour. Math. Phys. 12 (1971) 246.
[ 1441 F. Calogero, Jour. Math. Phys. 12 (1971) 419.
[ 1451 For a list of solvable potentials and other references in the field see M. Olshanetsky and A. Perelomov, Phys. Rep.
71 (1981) 314; ibid. 94 (1983) 6.
[ 1461 F. Haldane, Phys. Rev. Lett. 60 (1988) 635.
[ 1471 B. Shastry, Phys. Rev. Lett. 60 (1988) 639.
[ 1481 H. Frabm, Jour. Phys. A 26 (1993) L473.
[ 1491 A. Polychronakos, Phys. Rev. Lett. 69 ( 1992) 703.
[ 1501 D. Z. Freedman and P. F. Mende, Nucl. Phys. B 344 (1990) 317.
[ 1511 N. Froman and P. 0. Froman, JWKB Approximation, Contributions to the Theory, North Holland, Amsterdam
(1965).
[ 1521 J. B. Krieger and C. Rosenzweig, Phys. Rev. 164 (1967) 171; Jour. Chem. Phys. 47 (1967) 2942.
[ 1531 M. Seetharaman and S. Vasan, Jour. Phys. A 17 (1984) 2485,2493; ibid. A 18 (1985) 1041.
383
[ 1541 D. Barclay, A. Khare and U. Sukhatme, Phys. Lett. A 183 (1993) 263.
[ 1551
[ 1561
[ 1571
[ 1581
[ 1591
[ 1601
[ 1611
[ 1621
[ 1631
[ 1641
[ 1651
[ 1661
[ 1671
[ 1681
[ 1691
[ 1701
[ 1711
[ 1721
[ 1731
[ 1741
[ 1751
[ 1761
[ 1771
[ 1781
[ 1791
[ 1801
[ 1811
[ 1821
[ 1831
[ 1841
[ 1851
[ I861
[ 1871
[ 1881
[ 1891
[ 1901
[ 1911
[ 1921
[ 1931
[ 1941
[ 1951
[ 1961
[ 1971
[ 1981
[ 1991
[200]
[201]
[202]
[203]
384
[204]
[205]
[206]
[ 2071
[ 2081
[209]
[210]
[211]
[212]
[ 2131
[214]
[215]
[216]
[217]
[218]
[219]
[ 2201
[221]
[ 2221
[223]
[224]
[225]
[226]
[227]
[228]
[229]
[230]
[231]
[232]
[233]
[ 2341
[235]
[ 2361
[237]
[238]
[ 2391
[240]
[241]
[ 2421
[243]
[244]
[245]
[ 2461
[247]
[248]
385
[250]
[251]
[252]
[253]
[254]
[255]
[ 2561
[257]
[258]
[ 2591
[260]
[261]
[262]
[263]
[264]
[265]